]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
jbd2: fix block tag checksum verification brokenness
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
27dd4385 84static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
7f78e035 94MODULE_ALIAS_FS("ext2");
fa7614dd 95MODULE_ALIAS("ext2");
2035e776
TT		 96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
7f78e035 110MODULE_ALIAS_FS("ext3");
fa7614dd 111MODULE_ALIAS("ext3");
ba69f9ab
JK		 112#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113#else
114#define IS_EXT3_SB(sb) (0)
115#endif
bd81d8ee 116
d25425f8
DW		 117static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119{
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125}
126
a9c47317
DW		 127static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129{
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137}
138
139int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
141{
142 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147}
148
06db49e6 149void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 150{
06db49e6
TT		 151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
a9c47317
DW		 153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
155 return;
156
157 es->s_checksum = ext4_superblock_csum(sb, es);
158}
159
9933fc0a
TT		 160void *ext4_kvmalloc(size_t size, gfp_t flags)
161{
162 void *ret;
163
164 ret = kmalloc(size, flags);
165 if (!ret)
166 ret = __vmalloc(size, flags, PAGE_KERNEL);
167 return ret;
168}
169
170void *ext4_kvzalloc(size_t size, gfp_t flags)
171{
172 void *ret;
173
db9481c0 174 ret = kzalloc(size, flags);
9933fc0a
TT		 175 if (!ret)
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 return ret;
178}
179
180void ext4_kvfree(void *ptr)
181{
182 if (is_vmalloc_addr(ptr))
183 vfree(ptr);
184 else
185 kfree(ptr);
186
187}
188
8fadc143
AR		 189ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 struct ext4_group_desc *bg)
bd81d8ee 191{
3a14589c 192 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 194 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 195}
196
8fadc143
AR		 197ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 struct ext4_group_desc *bg)
bd81d8ee 199{
5272f837 200 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 202 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 203}
204
8fadc143
AR		 205ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 struct ext4_group_desc *bg)
bd81d8ee 207{
5272f837 208 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 210 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 211}
212
021b65bb
TT		 213__u32 ext4_free_group_clusters(struct super_block *sb,
214 struct ext4_group_desc *bg)
560671a0
AK		 215{
216 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 218 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 219}
220
221__u32 ext4_free_inodes_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
223{
224 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 226 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 227}
228
229__u32 ext4_used_dirs_count(struct super_block *sb,
230 struct ext4_group_desc *bg)
231{
232 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 234 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 235}
236
237__u32 ext4_itable_unused_count(struct super_block *sb,
238 struct ext4_group_desc *bg)
239{
240 return le16_to_cpu(bg->bg_itable_unused_lo) |
241 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 242 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 243}
244
8fadc143
AR		 245void ext4_block_bitmap_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 247{
3a14589c 248 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 251}
252
8fadc143
AR		 253void ext4_inode_bitmap_set(struct super_block *sb,
254 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 255{
5272f837 256 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 259}
260
8fadc143
AR		 261void ext4_inode_table_set(struct super_block *sb,
262 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 263{
5272f837 264 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 267}
268
021b65bb
TT		 269void ext4_free_group_clusters_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 271{
272 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
275}
276
277void ext4_free_inodes_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
279{
280 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
283}
284
285void ext4_used_dirs_set(struct super_block *sb,
286 struct ext4_group_desc *bg, __u32 count)
287{
288 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
291}
292
293void ext4_itable_unused_set(struct super_block *sb,
294 struct ext4_group_desc *bg, __u32 count)
295{
296 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299}
300
d3d1faf6 301
1c13d5c0
TT		 302static void __save_error_info(struct super_block *sb, const char *func,
303 unsigned int line)
304{
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306
307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309 es->s_last_error_time = cpu_to_le32(get_seconds());
310 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311 es->s_last_error_line = cpu_to_le32(line);
312 if (!es->s_first_error_time) {
313 es->s_first_error_time = es->s_last_error_time;
314 strncpy(es->s_first_error_func, func,
315 sizeof(es->s_first_error_func));
316 es->s_first_error_line = cpu_to_le32(line);
317 es->s_first_error_ino = es->s_last_error_ino;
318 es->s_first_error_block = es->s_last_error_block;
319 }
66e61a9e
TT		 320 /*
321 * Start the daily error reporting function if it hasn't been
322 * started already
323 */
324 if (!es->s_error_count)
325 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 326 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 327}
328
329static void save_error_info(struct super_block *sb, const char *func,
330 unsigned int line)
331{
332 __save_error_info(sb, func, line);
333 ext4_commit_super(sb, 1);
334}
335
7c2e7087
TT		 336/*
337 * The del_gendisk() function uninitializes the disk-specific data
338 * structures, including the bdi structure, without telling anyone
339 * else. Once this happens, any attempt to call mark_buffer_dirty()
340 * (for example, by ext4_commit_super), will cause a kernel OOPS.
341 * This is a kludge to prevent these oops until we can put in a proper
342 * hook in del_gendisk() to inform the VFS and file system layers.
343 */
344static int block_device_ejected(struct super_block *sb)
345{
346 struct inode *bd_inode = sb->s_bdev->bd_inode;
347 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348
349 return bdi->dev == NULL;
350}
351
18aadd47
BJ		 352static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353{
354 struct super_block *sb = journal->j_private;
355 struct ext4_sb_info *sbi = EXT4_SB(sb);
356 int error = is_journal_aborted(journal);
5d3ee208 357 struct ext4_journal_cb_entry *jce;
18aadd47 358
5d3ee208 359 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 360 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 361 while (!list_empty(&txn->t_private_list)) {
362 jce = list_entry(txn->t_private_list.next,
363 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 364 list_del_init(&jce->jce_list);
365 spin_unlock(&sbi->s_md_lock);
366 jce->jce_func(sb, jce, error);
367 spin_lock(&sbi->s_md_lock);
368 }
369 spin_unlock(&sbi->s_md_lock);
370}
1c13d5c0 371
ac27a0ec
DK		 372/* Deal with the reporting of failure conditions on a filesystem such as
373 * inconsistencies detected or read IO failures.
374 *
375 * On ext2, we can store the error state of the filesystem in the
617ba13b 376 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 377 * write ordering constraints on the superblock which prevent us from
378 * writing it out straight away; and given that the journal is about to
379 * be aborted, we can't rely on the current, or future, transactions to
380 * write out the superblock safely.
381 *
dab291af 382 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 383 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 384 * that error until we've noted it down and cleared it.
385 */
386
617ba13b 387static void ext4_handle_error(struct super_block *sb)
ac27a0ec 388{
ac27a0ec
DK		 389 if (sb->s_flags & MS_RDONLY)
390 return;
391
2b2d6d01 392 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 393 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 394
4ab2f15b 395 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 396 if (journal)
dab291af 397 jbd2_journal_abort(journal, -EIO);
ac27a0ec 398 }
2b2d6d01 399 if (test_opt(sb, ERRORS_RO)) {
b31e1552 400 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 401 sb->s_flags |= MS_RDONLY;
402 }
ac27a0ec 403 if (test_opt(sb, ERRORS_PANIC))
617ba13b 404 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 405 sb->s_id);
406}
407
12062ddd 408void __ext4_error(struct super_block *sb, const char *function,
c398eda0 409 unsigned int line, const char *fmt, ...)
ac27a0ec 410{
0ff2ea7d 411 struct va_format vaf;
ac27a0ec
DK		 412 va_list args;
413
414 va_start(args, fmt);
0ff2ea7d
JP		 415 vaf.fmt = fmt;
416 vaf.va = &args;
417 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
418 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 419 va_end(args);
f3fc0210 420 save_error_info(sb, function, line);
ac27a0ec 421
617ba13b 422 ext4_handle_error(sb);
ac27a0ec
DK		 423}
424
c398eda0
TT		 425void ext4_error_inode(struct inode *inode, const char *function,
426 unsigned int line, ext4_fsblk_t block,
273df556
FM		 427 const char *fmt, ...)
428{
429 va_list args;
f7c21177 430 struct va_format vaf;
1c13d5c0 431 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 432
1c13d5c0
TT		 433 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
434 es->s_last_error_block = cpu_to_le64(block);
435 save_error_info(inode->i_sb, function, line);
273df556 436 va_start(args, fmt);
f7c21177
TT		 437 vaf.fmt = fmt;
438 vaf.va = &args;
c398eda0 439 if (block)
d9ee81da
JP		 440 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
441 "inode #%lu: block %llu: comm %s: %pV\n",
442 inode->i_sb->s_id, function, line, inode->i_ino,
443 block, current->comm, &vaf);
444 else
445 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
446 "inode #%lu: comm %s: %pV\n",
447 inode->i_sb->s_id, function, line, inode->i_ino,
448 current->comm, &vaf);
273df556
FM		 449 va_end(args);
450
451 ext4_handle_error(inode->i_sb);
452}
453
c398eda0 454void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
273df556
FM		 457{
458 va_list args;
f7c21177 459 struct va_format vaf;
1c13d5c0 460 struct ext4_super_block *es;
496ad9aa 461 struct inode *inode = file_inode(file);
273df556
FM		 462 char pathname[80], *path;
463
1c13d5c0
TT		 464 es = EXT4_SB(inode->i_sb)->s_es;
465 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
466 save_error_info(inode->i_sb, function, line);
273df556 467 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 468 if (IS_ERR(path))
273df556 469 path = "(unknown)";
f7c21177
TT		 470 va_start(args, fmt);
471 vaf.fmt = fmt;
472 vaf.va = &args;
d9ee81da
JP		 473 if (block)
474 printk(KERN_CRIT
475 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
476 "block %llu: comm %s: path %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, path, &vaf);
479 else
480 printk(KERN_CRIT
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 current->comm, path, &vaf);
273df556
FM		 485 va_end(args);
486
487 ext4_handle_error(inode->i_sb);
488}
489
722887dd
TT		 490const char *ext4_decode_error(struct super_block *sb, int errno,
491 char nbuf[16])
ac27a0ec
DK		 492{
493 char *errstr = NULL;
494
495 switch (errno) {
496 case -EIO:
497 errstr = "IO failure";
498 break;
499 case -ENOMEM:
500 errstr = "Out of memory";
501 break;
502 case -EROFS:
78f1ddbb
TT		 503 if (!sb || (EXT4_SB(sb)->s_journal &&
504 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 505 errstr = "Journal has aborted";
506 else
507 errstr = "Readonly filesystem";
508 break;
509 default:
510 /* If the caller passed in an extra buffer for unknown
511 * errors, textualise them now. Else we just return
512 * NULL. */
513 if (nbuf) {
514 /* Check for truncated error codes... */
515 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
516 errstr = nbuf;
517 }
518 break;
519 }
520
521 return errstr;
522}
523
617ba13b 524/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 525 * automatically and invokes the appropriate error response. */
526
c398eda0
TT		 527void __ext4_std_error(struct super_block *sb, const char *function,
528 unsigned int line, int errno)
ac27a0ec
DK		 529{
530 char nbuf[16];
531 const char *errstr;
532
533 /* Special case: if the error is EROFS, and we're not already
534 * inside a transaction, then there's really no point in logging
535 * an error. */
536 if (errno == -EROFS && journal_current_handle() == NULL &&
537 (sb->s_flags & MS_RDONLY))
538 return;
539
617ba13b 540 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 541 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
542 sb->s_id, function, line, errstr);
1c13d5c0 543 save_error_info(sb, function, line);
ac27a0ec 544
617ba13b 545 ext4_handle_error(sb);
ac27a0ec
DK		 546}
547
548/*
617ba13b 549 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 550 * abort function may be used to deal with unrecoverable failures such
551 * as journal IO errors or ENOMEM at a critical moment in log management.
552 *
553 * We unconditionally force the filesystem into an ABORT|READONLY state,
554 * unless the error response on the fs has been set to panic in which
555 * case we take the easy way out and panic immediately.
556 */
557
c67d859e 558void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 559 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 560{
561 va_list args;
562
1c13d5c0 563 save_error_info(sb, function, line);
ac27a0ec 564 va_start(args, fmt);
c398eda0
TT		 565 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
566 function, line);
ac27a0ec
DK		 567 vprintk(fmt, args);
568 printk("\n");
569 va_end(args);
570
1c13d5c0
TT		 571 if ((sb->s_flags & MS_RDONLY) == 0) {
572 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
573 sb->s_flags |= MS_RDONLY;
574 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
575 if (EXT4_SB(sb)->s_journal)
576 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
577 save_error_info(sb, function, line);
578 }
ac27a0ec 579 if (test_opt(sb, ERRORS_PANIC))
617ba13b 580 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 581}
582
0ff2ea7d 583void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 584{
0ff2ea7d 585 struct va_format vaf;
b31e1552
ES		 586 va_list args;
587
588 va_start(args, fmt);
0ff2ea7d
JP		 589 vaf.fmt = fmt;
590 vaf.va = &args;
591 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 592 va_end(args);
593}
594
12062ddd 595void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 596 unsigned int line, const char *fmt, ...)
ac27a0ec 597{
0ff2ea7d 598 struct va_format vaf;
ac27a0ec
DK		 599 va_list args;
600
601 va_start(args, fmt);
0ff2ea7d
JP		 602 vaf.fmt = fmt;
603 vaf.va = &args;
604 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
605 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 606 va_end(args);
607}
608
e29136f8
TT		 609void __ext4_grp_locked_error(const char *function, unsigned int line,
610 struct super_block *sb, ext4_group_t grp,
611 unsigned long ino, ext4_fsblk_t block,
612 const char *fmt, ...)
5d1b1b3f
AK		 613__releases(bitlock)
614__acquires(bitlock)
615{
0ff2ea7d 616 struct va_format vaf;
5d1b1b3f
AK		 617 va_list args;
618 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
619
1c13d5c0
TT		 620 es->s_last_error_ino = cpu_to_le32(ino);
621 es->s_last_error_block = cpu_to_le64(block);
622 __save_error_info(sb, function, line);
0ff2ea7d 623
5d1b1b3f 624 va_start(args, fmt);
0ff2ea7d
JP		 625
626 vaf.fmt = fmt;
627 vaf.va = &args;
21149d61 628 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 629 sb->s_id, function, line, grp);
630 if (ino)
0ff2ea7d 631 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 632 if (block)
0ff2ea7d
JP		 633 printk(KERN_CONT "block %llu:", (unsigned long long) block);
634 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 635 va_end(args);
636
637 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 638 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 639 return;
640 }
1c13d5c0 641
5d1b1b3f
AK		 642 ext4_unlock_group(sb, grp);
643 ext4_handle_error(sb);
644 /*
645 * We only get here in the ERRORS_RO case; relocking the group
646 * may be dangerous, but nothing bad will happen since the
647 * filesystem will have already been marked read/only and the
648 * journal has been aborted. We return 1 as a hint to callers
649 * who might what to use the return value from
25985edc 650 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 651 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
652 * aggressively from the ext4 function in question, with a
653 * more appropriate error code.
654 */
655 ext4_lock_group(sb, grp);
656 return;
657}
658
617ba13b 659void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 660{
617ba13b 661 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 662
617ba13b 663 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 664 return;
665
12062ddd 666 ext4_warning(sb,
ac27a0ec
DK		 667 "updating to rev %d because of new feature flag, "
668 "running e2fsck is recommended",
617ba13b 669 EXT4_DYNAMIC_REV);
ac27a0ec 670
617ba13b
MC		 671 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
672 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
673 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 674 /* leave es->s_feature_*compat flags alone */
675 /* es->s_uuid will be set by e2fsck if empty */
676
677 /*
678 * The rest of the superblock fields should be zero, and if not it
679 * means they are likely already in use, so leave them alone. We
680 * can leave it up to e2fsck to clean up any inconsistencies there.
681 */
682}
683
684/*
685 * Open the external journal device
686 */
b31e1552 687static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 688{
689 struct block_device *bdev;
690 char b[BDEVNAME_SIZE];
691
d4d77629 692 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 693 if (IS_ERR(bdev))
694 goto fail;
695 return bdev;
696
697fail:
b31e1552 698 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 699 __bdevname(dev, b), PTR_ERR(bdev));
700 return NULL;
701}
702
703/*
704 * Release the journal device
705 */
4385bab1 706static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 707{
4385bab1 708 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 709}
710
4385bab1 711static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 712{
713 struct block_device *bdev;
ac27a0ec
DK		 714 bdev = sbi->journal_bdev;
715 if (bdev) {
4385bab1 716 ext4_blkdev_put(bdev);
ac27a0ec
DK		 717 sbi->journal_bdev = NULL;
718 }
ac27a0ec
DK		 719}
720
721static inline struct inode *orphan_list_entry(struct list_head *l)
722{
617ba13b 723 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 724}
725
617ba13b 726static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 727{
728 struct list_head *l;
729
b31e1552
ES		 730 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
731 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 732
733 printk(KERN_ERR "sb_info orphan list:\n");
734 list_for_each(l, &sbi->s_orphan) {
735 struct inode *inode = orphan_list_entry(l);
736 printk(KERN_ERR " "
737 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
738 inode->i_sb->s_id, inode->i_ino, inode,
739 inode->i_mode, inode->i_nlink,
740 NEXT_ORPHAN(inode));
741 }
742}
743
2b2d6d01 744static void ext4_put_super(struct super_block *sb)
ac27a0ec 745{
617ba13b
MC		 746 struct ext4_sb_info *sbi = EXT4_SB(sb);
747 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 748 int i, err;
ac27a0ec 749
857ac889 750 ext4_unregister_li_request(sb);
e0ccfd95
CH		 751 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
752
4c0425ff
MC		 753 flush_workqueue(sbi->dio_unwritten_wq);
754 destroy_workqueue(sbi->dio_unwritten_wq);
755
0390131b
FM		 756 if (sbi->s_journal) {
757 err = jbd2_journal_destroy(sbi->s_journal);
758 sbi->s_journal = NULL;
759 if (err < 0)
c67d859e 760 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 761 }
d4edac31 762
74cd15cd 763 ext4_es_unregister_shrinker(sb);
a1c6c569 764 del_timer(&sbi->s_err_report);
d4edac31
JB		 765 ext4_release_system_zone(sb);
766 ext4_mb_release(sb);
767 ext4_ext_release(sb);
768 ext4_xattr_put_super(sb);
769
ac27a0ec 770 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 771 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 772 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 773 }
58c5873a 774 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 775 ext4_commit_super(sb, 1);
776
240799cd 777 if (sbi->s_proc) {
66acdcf4 778 remove_proc_entry("options", sbi->s_proc);
9f6200bb 779 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 780 }
3197ebdb 781 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 782
783 for (i = 0; i < sbi->s_gdb_count; i++)
784 brelse(sbi->s_group_desc[i]);
f18a5f21 785 ext4_kvfree(sbi->s_group_desc);
9933fc0a 786 ext4_kvfree(sbi->s_flex_groups);
57042651 787 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 788 percpu_counter_destroy(&sbi->s_freeinodes_counter);
789 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 790 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 791 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
ac27a0ec
DK		 792 brelse(sbi->s_sbh);
793#ifdef CONFIG_QUOTA
794 for (i = 0; i < MAXQUOTAS; i++)
795 kfree(sbi->s_qf_names[i]);
796#endif
797
798 /* Debugging code just in case the in-memory inode orphan list
799 * isn't empty. The on-disk one can be non-empty if we've
800 * detected an error and taken the fs readonly, but the
801 * in-memory list had better be clean by this point. */
802 if (!list_empty(&sbi->s_orphan))
803 dump_orphan_list(sb, sbi);
804 J_ASSERT(list_empty(&sbi->s_orphan));
805
f98393a6 806 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 807 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
808 /*
809 * Invalidate the journal device's buffers. We don't want them
810 * floating about in memory - the physical journal device may
811 * hotswapped, and it breaks the `ro-after' testing code.
812 */
813 sync_blockdev(sbi->journal_bdev);
f98393a6 814 invalidate_bdev(sbi->journal_bdev);
617ba13b 815 ext4_blkdev_remove(sbi);
ac27a0ec 816 }
c5e06d10
JL		 817 if (sbi->s_mmp_tsk)
818 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 819 sb->s_fs_info = NULL;
3197ebdb
TT		 820 /*
821 * Now that we are completely done shutting down the
822 * superblock, we need to actually destroy the kobject.
823 */
3197ebdb
TT		 824 kobject_put(&sbi->s_kobj);
825 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 826 if (sbi->s_chksum_driver)
827 crypto_free_shash(sbi->s_chksum_driver);
705895b6 828 kfree(sbi->s_blockgroup_lock);
ac27a0ec 829 kfree(sbi);
ac27a0ec
DK		 830}
831
e18b890b 832static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 833
834/*
835 * Called inside transaction, so use GFP_NOFS
836 */
617ba13b 837static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 838{
617ba13b 839 struct ext4_inode_info *ei;
ac27a0ec 840
e6b4f8da 841 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 842 if (!ei)
843 return NULL;
0b8e58a1 844
ac27a0ec 845 ei->vfs_inode.i_version = 1;
c9de560d
AT		 846 INIT_LIST_HEAD(&ei->i_prealloc_list);
847 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 848 ext4_es_init_tree(&ei->i_es_tree);
849 rwlock_init(&ei->i_es_lock);
74cd15cd
ZL		 850 INIT_LIST_HEAD(&ei->i_es_lru);
851 ei->i_es_lru_nr = 0;
d2a17637
MC		 852 ei->i_reserved_data_blocks = 0;
853 ei->i_reserved_meta_blocks = 0;
854 ei->i_allocated_meta_blocks = 0;
9d0be502 855 ei->i_da_metadata_calc_len = 0;
7e731bc9 856 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 857 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 858#ifdef CONFIG_QUOTA
859 ei->i_reserved_quota = 0;
860#endif
8aefcd55 861 ei->jinode = NULL;
c7064ef1 862 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 863 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 864 ei->i_sync_tid = 0;
865 ei->i_datasync_tid = 0;
f7ad6d2e 866 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 867 atomic_set(&ei->i_unwritten, 0);
84c17543 868 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
0b8e58a1 869
ac27a0ec
DK		 870 return &ei->vfs_inode;
871}
872
7ff9c073
TT		 873static int ext4_drop_inode(struct inode *inode)
874{
875 int drop = generic_drop_inode(inode);
876
877 trace_ext4_drop_inode(inode, drop);
878 return drop;
879}
880
fa0d7e3d
NP		 881static void ext4_i_callback(struct rcu_head *head)
882{
883 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 884 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
885}
886
617ba13b 887static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 888{
9f7dd93d 889 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 890 ext4_msg(inode->i_sb, KERN_ERR,
891 "Inode %lu (%p): orphan list check failed!",
892 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 893 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
894 EXT4_I(inode), sizeof(struct ext4_inode_info),
895 true);
896 dump_stack();
897 }
fa0d7e3d 898 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 899}
900
51cc5068 901static void init_once(void *foo)
ac27a0ec 902{
617ba13b 903 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 904
a35afb83 905 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 906 init_rwsem(&ei->xattr_sem);
0e855ac8 907 init_rwsem(&ei->i_data_sem);
a35afb83 908 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 909}
910
911static int init_inodecache(void)
912{
617ba13b
MC		 913 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
914 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 915 0, (SLAB_RECLAIM_ACCOUNT|
916 SLAB_MEM_SPREAD),
20c2df83 917 init_once);
617ba13b 918 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 919 return -ENOMEM;
920 return 0;
921}
922
923static void destroy_inodecache(void)
924{
8c0a8537
KS		 925 /*
926 * Make sure all delayed rcu free inodes are flushed before we
927 * destroy cache.
928 */
929 rcu_barrier();
617ba13b 930 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 931}
932
0930fcc1 933void ext4_clear_inode(struct inode *inode)
ac27a0ec 934{
0930fcc1 935 invalidate_inode_buffers(inode);
dbd5768f 936 clear_inode(inode);
9f754758 937 dquot_drop(inode);
c2ea3fde 938 ext4_discard_preallocations(inode);
51865fda 939 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 940 ext4_es_lru_del(inode);
8aefcd55
TT		 941 if (EXT4_I(inode)->jinode) {
942 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
943 EXT4_I(inode)->jinode);
944 jbd2_free_inode(EXT4_I(inode)->jinode);
945 EXT4_I(inode)->jinode = NULL;
946 }
ac27a0ec
DK		 947}
948
1b961ac0 949static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 950 u64 ino, u32 generation)
ac27a0ec 951{
ac27a0ec 952 struct inode *inode;
ac27a0ec 953
617ba13b 954 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 955 return ERR_PTR(-ESTALE);
617ba13b 956 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 957 return ERR_PTR(-ESTALE);
958
959 /* iget isn't really right if the inode is currently unallocated!!
960 *
617ba13b 961 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 962 * deleted, so we should be safe.
963 *
964 * Currently we don't know the generation for parent directory, so
965 * a generation of 0 means "accept any"
966 */
1d1fe1ee
DH		 967 inode = ext4_iget(sb, ino);
968 if (IS_ERR(inode))
969 return ERR_CAST(inode);
970 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 971 iput(inode);
972 return ERR_PTR(-ESTALE);
973 }
1b961ac0
CH		 974
975 return inode;
976}
977
978static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 979 int fh_len, int fh_type)
1b961ac0
CH		 980{
981 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
982 ext4_nfs_get_inode);
983}
984
985static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 986 int fh_len, int fh_type)
1b961ac0
CH		 987{
988 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
989 ext4_nfs_get_inode);
ac27a0ec
DK		 990}
991
c39a7f84
TO		 992/*
993 * Try to release metadata pages (indirect blocks, directories) which are
994 * mapped via the block device. Since these pages could have journal heads
995 * which would prevent try_to_free_buffers() from freeing them, we must use
996 * jbd2 layer's try_to_free_buffers() function to release them.
997 */
0b8e58a1
AD		 998static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
999 gfp_t wait)
c39a7f84
TO		 1000{
1001 journal_t *journal = EXT4_SB(sb)->s_journal;
1002
1003 WARN_ON(PageChecked(page));
1004 if (!page_has_buffers(page))
1005 return 0;
1006 if (journal)
1007 return jbd2_journal_try_to_free_buffers(journal, page,
1008 wait & ~__GFP_WAIT);
1009 return try_to_free_buffers(page);
1010}
1011
ac27a0ec 1012#ifdef CONFIG_QUOTA
af5bc92d 1013#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1014#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1015
617ba13b
MC		 1016static int ext4_write_dquot(struct dquot *dquot);
1017static int ext4_acquire_dquot(struct dquot *dquot);
1018static int ext4_release_dquot(struct dquot *dquot);
1019static int ext4_mark_dquot_dirty(struct dquot *dquot);
1020static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1021static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1022 struct path *path);
7c319d32
AK		 1023static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1024 int format_id);
ca0e05e4 1025static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1026static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1027static int ext4_quota_on_mount(struct super_block *sb, int type);
1028static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1029 size_t len, loff_t off);
617ba13b 1030static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1031 const char *data, size_t len, loff_t off);
7c319d32
AK		 1032static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1033 unsigned int flags);
1034static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1035
61e225dc 1036static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1037 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1038 .write_dquot = ext4_write_dquot,
1039 .acquire_dquot = ext4_acquire_dquot,
1040 .release_dquot = ext4_release_dquot,
1041 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1042 .write_info = ext4_write_info,
1043 .alloc_dquot = dquot_alloc,
1044 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1045};
1046
0d54b217 1047static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1048 .quota_on = ext4_quota_on,
ca0e05e4 1049 .quota_off = ext4_quota_off,
287a8095
CH		 1050 .quota_sync = dquot_quota_sync,
1051 .get_info = dquot_get_dqinfo,
1052 .set_info = dquot_set_dqinfo,
1053 .get_dqblk = dquot_get_dqblk,
1054 .set_dqblk = dquot_set_dqblk
ac27a0ec 1055};
7c319d32
AK		 1056
1057static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1058 .quota_on_meta = ext4_quota_on_sysfile,
1059 .quota_off = ext4_quota_off_sysfile,
1060 .quota_sync = dquot_quota_sync,
1061 .get_info = dquot_get_dqinfo,
1062 .set_info = dquot_set_dqinfo,
1063 .get_dqblk = dquot_get_dqblk,
1064 .set_dqblk = dquot_set_dqblk
1065};
ac27a0ec
DK		 1066#endif
1067
ee9b6d61 1068static const struct super_operations ext4_sops = {
617ba13b
MC		 1069 .alloc_inode = ext4_alloc_inode,
1070 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1071 .write_inode = ext4_write_inode,
1072 .dirty_inode = ext4_dirty_inode,
7ff9c073 1073 .drop_inode = ext4_drop_inode,
0930fcc1 1074 .evict_inode = ext4_evict_inode,
617ba13b 1075 .put_super = ext4_put_super,
617ba13b 1076 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1077 .freeze_fs = ext4_freeze,
1078 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1079 .statfs = ext4_statfs,
1080 .remount_fs = ext4_remount,
617ba13b 1081 .show_options = ext4_show_options,
ac27a0ec 1082#ifdef CONFIG_QUOTA
617ba13b
MC		 1083 .quota_read = ext4_quota_read,
1084 .quota_write = ext4_quota_write,
ac27a0ec 1085#endif
c39a7f84 1086 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1087};
1088
9ca92389
TT		 1089static const struct super_operations ext4_nojournal_sops = {
1090 .alloc_inode = ext4_alloc_inode,
1091 .destroy_inode = ext4_destroy_inode,
1092 .write_inode = ext4_write_inode,
1093 .dirty_inode = ext4_dirty_inode,
7ff9c073 1094 .drop_inode = ext4_drop_inode,
0930fcc1 1095 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1096 .put_super = ext4_put_super,
1097 .statfs = ext4_statfs,
1098 .remount_fs = ext4_remount,
9ca92389
TT		 1099 .show_options = ext4_show_options,
1100#ifdef CONFIG_QUOTA
1101 .quota_read = ext4_quota_read,
1102 .quota_write = ext4_quota_write,
1103#endif
1104 .bdev_try_to_free_page = bdev_try_to_free_page,
1105};
1106
39655164 1107static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1108 .fh_to_dentry = ext4_fh_to_dentry,
1109 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1110 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1111};
1112
1113enum {
1114 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1115 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1116 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1117 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1118 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1119 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1120 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1121 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1122 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1123 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1124 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1125 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1126 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1127 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1128 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1129 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1130 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1131 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1132 Opt_max_dir_size_kb,
ac27a0ec
DK		 1133};
1134
a447c093 1135static const match_table_t tokens = {
ac27a0ec
DK		 1136 {Opt_bsd_df, "bsddf"},
1137 {Opt_minix_df, "minixdf"},
1138 {Opt_grpid, "grpid"},
1139 {Opt_grpid, "bsdgroups"},
1140 {Opt_nogrpid, "nogrpid"},
1141 {Opt_nogrpid, "sysvgroups"},
1142 {Opt_resgid, "resgid=%u"},
1143 {Opt_resuid, "resuid=%u"},
1144 {Opt_sb, "sb=%u"},
1145 {Opt_err_cont, "errors=continue"},
1146 {Opt_err_panic, "errors=panic"},
1147 {Opt_err_ro, "errors=remount-ro"},
1148 {Opt_nouid32, "nouid32"},
ac27a0ec 1149 {Opt_debug, "debug"},
72578c33
TT		 1150 {Opt_removed, "oldalloc"},
1151 {Opt_removed, "orlov"},
ac27a0ec
DK		 1152 {Opt_user_xattr, "user_xattr"},
1153 {Opt_nouser_xattr, "nouser_xattr"},
1154 {Opt_acl, "acl"},
1155 {Opt_noacl, "noacl"},
e3bb52ae 1156 {Opt_noload, "norecovery"},
5a916be1 1157 {Opt_noload, "noload"},
72578c33
TT		 1158 {Opt_removed, "nobh"},
1159 {Opt_removed, "bh"},
ac27a0ec 1160 {Opt_commit, "commit=%u"},
30773840
TT		 1161 {Opt_min_batch_time, "min_batch_time=%u"},
1162 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1163 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1164 {Opt_journal_checksum, "journal_checksum"},
1165 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1166 {Opt_abort, "abort"},
1167 {Opt_data_journal, "data=journal"},
1168 {Opt_data_ordered, "data=ordered"},
1169 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1170 {Opt_data_err_abort, "data_err=abort"},
1171 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1172 {Opt_offusrjquota, "usrjquota="},
1173 {Opt_usrjquota, "usrjquota=%s"},
1174 {Opt_offgrpjquota, "grpjquota="},
1175 {Opt_grpjquota, "grpjquota=%s"},
1176 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1177 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1178 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1179 {Opt_grpquota, "grpquota"},
1180 {Opt_noquota, "noquota"},
1181 {Opt_quota, "quota"},
1182 {Opt_usrquota, "usrquota"},
1183 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1184 {Opt_barrier, "barrier"},
1185 {Opt_nobarrier, "nobarrier"},
25ec56b5 1186 {Opt_i_version, "i_version"},
c9de560d 1187 {Opt_stripe, "stripe=%u"},
64769240 1188 {Opt_delalloc, "delalloc"},
dd919b98 1189 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1190 {Opt_removed, "mblk_io_submit"},
1191 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1192 {Opt_block_validity, "block_validity"},
1193 {Opt_noblock_validity, "noblock_validity"},
240799cd 1194 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1195 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1196 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1197 {Opt_auto_da_alloc, "auto_da_alloc"},
1198 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1199 {Opt_dioread_nolock, "dioread_nolock"},
1200 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1201 {Opt_discard, "discard"},
1202 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1203 {Opt_init_itable, "init_itable=%u"},
1204 {Opt_init_itable, "init_itable"},
1205 {Opt_noinit_itable, "noinit_itable"},
df981d03 1206 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1207 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1208 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1209 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1210 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1211 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1212 {Opt_err, NULL},
ac27a0ec
DK		 1213};
1214
617ba13b 1215static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1216{
617ba13b 1217 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1218 char *options = (char *) *data;
1219
1220 if (!options || strncmp(options, "sb=", 3) != 0)
1221 return 1; /* Default location */
0b8e58a1 1222
ac27a0ec 1223 options += 3;
0b8e58a1 1224 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1225 sb_block = simple_strtoul(options, &options, 0);
1226 if (*options && *options != ',') {
4776004f 1227 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1228 (char *) *data);
1229 return 1;
1230 }
1231 if (*options == ',')
1232 options++;
1233 *data = (void *) options;
0b8e58a1 1234
ac27a0ec
DK		 1235 return sb_block;
1236}
1237
b3881f74 1238#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1239static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1240 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1241
56c50f11
DM		 1242#ifdef CONFIG_QUOTA
1243static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1244{
1245 struct ext4_sb_info *sbi = EXT4_SB(sb);
1246 char *qname;
03dafb5f 1247 int ret = -1;
56c50f11
DM		 1248
1249 if (sb_any_quota_loaded(sb) &&
1250 !sbi->s_qf_names[qtype]) {
1251 ext4_msg(sb, KERN_ERR,
1252 "Cannot change journaled "
1253 "quota options when quota turned on");
57f73c2c 1254 return -1;
56c50f11 1255 }
262b4662
JK		 1256 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1257 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1258 "when QUOTA feature is enabled");
1259 return -1;
1260 }
56c50f11
DM		 1261 qname = match_strdup(args);
1262 if (!qname) {
1263 ext4_msg(sb, KERN_ERR,
1264 "Not enough memory for storing quotafile name");
57f73c2c 1265 return -1;
56c50f11 1266 }
03dafb5f
CG		 1267 if (sbi->s_qf_names[qtype]) {
1268 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1269 ret = 1;
1270 else
1271 ext4_msg(sb, KERN_ERR,
1272 "%s quota file already specified",
1273 QTYPE2NAME(qtype));
1274 goto errout;
56c50f11 1275 }
03dafb5f 1276 if (strchr(qname, '/')) {
56c50f11
DM		 1277 ext4_msg(sb, KERN_ERR,
1278 "quotafile must be on filesystem root");
03dafb5f 1279 goto errout;
56c50f11 1280 }
03dafb5f 1281 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1282 set_opt(sb, QUOTA);
56c50f11 1283 return 1;
03dafb5f
CG		 1284errout:
1285 kfree(qname);
1286 return ret;
56c50f11
DM		 1287}
1288
1289static int clear_qf_name(struct super_block *sb, int qtype)
1290{
1291
1292 struct ext4_sb_info *sbi = EXT4_SB(sb);
1293
1294 if (sb_any_quota_loaded(sb) &&
1295 sbi->s_qf_names[qtype]) {
1296 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1297 " when quota turned on");
57f73c2c 1298 return -1;
56c50f11 1299 }
03dafb5f 1300 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1301 sbi->s_qf_names[qtype] = NULL;
1302 return 1;
1303}
1304#endif
1305
26092bf5
TT		 1306#define MOPT_SET 0x0001
1307#define MOPT_CLEAR 0x0002
1308#define MOPT_NOSUPPORT 0x0004
1309#define MOPT_EXPLICIT 0x0008
1310#define MOPT_CLEAR_ERR 0x0010
1311#define MOPT_GTE0 0x0020
ac27a0ec 1312#ifdef CONFIG_QUOTA
26092bf5
TT		 1313#define MOPT_Q 0
1314#define MOPT_QFMT 0x0040
1315#else
1316#define MOPT_Q MOPT_NOSUPPORT
1317#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1318#endif
26092bf5 1319#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1320#define MOPT_NO_EXT2 0x0100
1321#define MOPT_NO_EXT3 0x0200
1322#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1323
1324static const struct mount_opts {
1325 int token;
1326 int mount_opt;
1327 int flags;
1328} ext4_mount_opts[] = {
1329 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1330 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1331 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1332 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1333 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1334 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1335 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1336 MOPT_EXT4_ONLY | MOPT_SET},
1337 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1338 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1339 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1340 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1341 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1342 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1343 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1344 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1345 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1346 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1347 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1348 EXT4_MOUNT_JOURNAL_CHECKSUM),
1349 MOPT_EXT4_ONLY | MOPT_SET},
1350 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1351 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1352 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1353 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1354 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1355 MOPT_NO_EXT2 | MOPT_SET},
1356 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1357 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1358 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1359 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1360 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1361 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1362 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1363 {Opt_commit, 0, MOPT_GTE0},
1364 {Opt_max_batch_time, 0, MOPT_GTE0},
1365 {Opt_min_batch_time, 0, MOPT_GTE0},
1366 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1367 {Opt_init_itable, 0, MOPT_GTE0},
1368 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1369 {Opt_resuid, 0, MOPT_GTE0},
1370 {Opt_resgid, 0, MOPT_GTE0},
1371 {Opt_journal_dev, 0, MOPT_GTE0},
1372 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1373 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1375 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1376 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1377 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1378 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1379#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1380 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1381 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1382#else
26092bf5
TT		 1383 {Opt_acl, 0, MOPT_NOSUPPORT},
1384 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1385#endif
26092bf5
TT		 1386 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1387 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1388 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1389 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1390 MOPT_SET | MOPT_Q},
1391 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1392 MOPT_SET | MOPT_Q},
1393 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1394 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1395 {Opt_usrjquota, 0, MOPT_Q},
1396 {Opt_grpjquota, 0, MOPT_Q},
1397 {Opt_offusrjquota, 0, MOPT_Q},
1398 {Opt_offgrpjquota, 0, MOPT_Q},
1399 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1400 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1401 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1402 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1403 {Opt_err, 0, 0}
1404};
1405
1406static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1407 substring_t *args, unsigned long *journal_devnum,
1408 unsigned int *journal_ioprio, int is_remount)
1409{
1410 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 const struct mount_opts *m;
08cefc7a
EB		 1412 kuid_t uid;
1413 kgid_t gid;
26092bf5
TT		 1414 int arg = 0;
1415
57f73c2c
TT		 1416#ifdef CONFIG_QUOTA
1417 if (token == Opt_usrjquota)
1418 return set_qf_name(sb, USRQUOTA, &args[0]);
1419 else if (token == Opt_grpjquota)
1420 return set_qf_name(sb, GRPQUOTA, &args[0]);
1421 else if (token == Opt_offusrjquota)
1422 return clear_qf_name(sb, USRQUOTA);
1423 else if (token == Opt_offgrpjquota)
1424 return clear_qf_name(sb, GRPQUOTA);
1425#endif
26092bf5 1426 switch (token) {
f7048605
TT		 1427 case Opt_noacl:
1428 case Opt_nouser_xattr:
1429 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1430 break;
26092bf5
TT		 1431 case Opt_sb:
1432 return 1; /* handled by get_sb_block() */
1433 case Opt_removed:
5f3633e3 1434 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1435 return 1;
26092bf5
TT		 1436 case Opt_abort:
1437 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1438 return 1;
1439 case Opt_i_version:
1440 sb->s_flags |= MS_I_VERSION;
1441 return 1;
26092bf5
TT		 1442 }
1443
5f3633e3
JK		 1444 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1445 if (token == m->token)
1446 break;
1447
1448 if (m->token == Opt_err) {
1449 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1450 "or missing value", opt);
1451 return -1;
1452 }
1453
8dc0aa8c
TT		 1454 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1455 ext4_msg(sb, KERN_ERR,
1456 "Mount option \"%s\" incompatible with ext2", opt);
1457 return -1;
1458 }
1459 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1460 ext4_msg(sb, KERN_ERR,
1461 "Mount option \"%s\" incompatible with ext3", opt);
1462 return -1;
1463 }
1464
5f3633e3
JK		 1465 if (args->from && match_int(args, &arg))
1466 return -1;
1467 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1468 return -1;
1469 if (m->flags & MOPT_EXPLICIT)
1470 set_opt2(sb, EXPLICIT_DELALLOC);
1471 if (m->flags & MOPT_CLEAR_ERR)
1472 clear_opt(sb, ERRORS_MASK);
1473 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1474 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1475 "options when quota turned on");
1476 return -1;
1477 }
1478
1479 if (m->flags & MOPT_NOSUPPORT) {
1480 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1481 } else if (token == Opt_commit) {
1482 if (arg == 0)
1483 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1484 sbi->s_commit_interval = HZ * arg;
1485 } else if (token == Opt_max_batch_time) {
1486 if (arg == 0)
1487 arg = EXT4_DEF_MAX_BATCH_TIME;
1488 sbi->s_max_batch_time = arg;
1489 } else if (token == Opt_min_batch_time) {
1490 sbi->s_min_batch_time = arg;
1491 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1492 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1493 ext4_msg(sb, KERN_ERR,
1494 "EXT4-fs: inode_readahead_blks must be "
1495 "0 or a power of 2 smaller than 2^31");
26092bf5 1496 return -1;
5f3633e3
JK		 1497 }
1498 sbi->s_inode_readahead_blks = arg;
1499 } else if (token == Opt_init_itable) {
1500 set_opt(sb, INIT_INODE_TABLE);
1501 if (!args->from)
1502 arg = EXT4_DEF_LI_WAIT_MULT;
1503 sbi->s_li_wait_mult = arg;
1504 } else if (token == Opt_max_dir_size_kb) {
1505 sbi->s_max_dir_size_kb = arg;
1506 } else if (token == Opt_stripe) {
1507 sbi->s_stripe = arg;
1508 } else if (token == Opt_resuid) {
1509 uid = make_kuid(current_user_ns(), arg);
1510 if (!uid_valid(uid)) {
1511 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1512 return -1;
1513 }
5f3633e3
JK		 1514 sbi->s_resuid = uid;
1515 } else if (token == Opt_resgid) {
1516 gid = make_kgid(current_user_ns(), arg);
1517 if (!gid_valid(gid)) {
1518 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1519 return -1;
1520 }
1521 sbi->s_resgid = gid;
1522 } else if (token == Opt_journal_dev) {
1523 if (is_remount) {
1524 ext4_msg(sb, KERN_ERR,
1525 "Cannot specify journal on remount");
1526 return -1;
1527 }
1528 *journal_devnum = arg;
1529 } else if (token == Opt_journal_ioprio) {
1530 if (arg > 7) {
1531 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1532 " (must be 0-7)");
1533 return -1;
1534 }
1535 *journal_ioprio =
1536 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1537 } else if (m->flags & MOPT_DATAJ) {
1538 if (is_remount) {
1539 if (!sbi->s_journal)
1540 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1541 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1542 ext4_msg(sb, KERN_ERR,
26092bf5 1543 "Cannot change data mode on remount");
26092bf5 1544 return -1;
ac27a0ec 1545 }
26092bf5 1546 } else {
5f3633e3
JK		 1547 clear_opt(sb, DATA_FLAGS);
1548 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1549 }
5f3633e3
JK		 1550#ifdef CONFIG_QUOTA
1551 } else if (m->flags & MOPT_QFMT) {
1552 if (sb_any_quota_loaded(sb) &&
1553 sbi->s_jquota_fmt != m->mount_opt) {
1554 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1555 "quota options when quota turned on");
1556 return -1;
1557 }
262b4662
JK		 1558 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1559 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1560 ext4_msg(sb, KERN_ERR,
1561 "Cannot set journaled quota options "
1562 "when QUOTA feature is enabled");
1563 return -1;
1564 }
5f3633e3
JK		 1565 sbi->s_jquota_fmt = m->mount_opt;
1566#endif
1567 } else {
1568 if (!args->from)
1569 arg = 1;
1570 if (m->flags & MOPT_CLEAR)
1571 arg = !arg;
1572 else if (unlikely(!(m->flags & MOPT_SET))) {
1573 ext4_msg(sb, KERN_WARNING,
1574 "buggy handling of option %s", opt);
1575 WARN_ON(1);
1576 return -1;
1577 }
1578 if (arg != 0)
1579 sbi->s_mount_opt |= m->mount_opt;
1580 else
1581 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1582 }
5f3633e3 1583 return 1;
26092bf5
TT		 1584}
1585
1586static int parse_options(char *options, struct super_block *sb,
1587 unsigned long *journal_devnum,
1588 unsigned int *journal_ioprio,
1589 int is_remount)
1590{
1591 struct ext4_sb_info *sbi = EXT4_SB(sb);
1592 char *p;
1593 substring_t args[MAX_OPT_ARGS];
1594 int token;
1595
1596 if (!options)
1597 return 1;
1598
1599 while ((p = strsep(&options, ",")) != NULL) {
1600 if (!*p)
1601 continue;
1602 /*
1603 * Initialize args struct so we know whether arg was
1604 * found; some options take optional arguments.
1605 */
caecd0af 1606 args[0].to = args[0].from = NULL;
26092bf5
TT		 1607 token = match_token(p, tokens, args);
1608 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1609 journal_ioprio, is_remount) < 0)
1610 return 0;
ac27a0ec
DK		 1611 }
1612#ifdef CONFIG_QUOTA
262b4662
JK		 1613 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1614 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1615 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1616 "feature is enabled");
1617 return 0;
1618 }
ac27a0ec 1619 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1620 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1621 clear_opt(sb, USRQUOTA);
ac27a0ec 1622
482a7425 1623 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1624 clear_opt(sb, GRPQUOTA);
ac27a0ec 1625
56c50f11 1626 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1627 ext4_msg(sb, KERN_ERR, "old and new quota "
1628 "format mixing");
ac27a0ec
DK		 1629 return 0;
1630 }
1631
1632 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1633 ext4_msg(sb, KERN_ERR, "journaled quota format "
1634 "not specified");
ac27a0ec
DK		 1635 return 0;
1636 }
1637 } else {
1638 if (sbi->s_jquota_fmt) {
b31e1552 1639 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1640 "specified with no journaling "
b31e1552 1641 "enabled");
ac27a0ec
DK		 1642 return 0;
1643 }
1644 }
1645#endif
261cb20c
JK		 1646 if (test_opt(sb, DIOREAD_NOLOCK)) {
1647 int blocksize =
1648 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1649
1650 if (blocksize < PAGE_CACHE_SIZE) {
1651 ext4_msg(sb, KERN_ERR, "can't mount with "
1652 "dioread_nolock if block size != PAGE_SIZE");
1653 return 0;
1654 }
1655 }
ac27a0ec
DK		 1656 return 1;
1657}
1658
2adf6da8
TT		 1659static inline void ext4_show_quota_options(struct seq_file *seq,
1660 struct super_block *sb)
1661{
1662#if defined(CONFIG_QUOTA)
1663 struct ext4_sb_info *sbi = EXT4_SB(sb);
1664
1665 if (sbi->s_jquota_fmt) {
1666 char *fmtname = "";
1667
1668 switch (sbi->s_jquota_fmt) {
1669 case QFMT_VFS_OLD:
1670 fmtname = "vfsold";
1671 break;
1672 case QFMT_VFS_V0:
1673 fmtname = "vfsv0";
1674 break;
1675 case QFMT_VFS_V1:
1676 fmtname = "vfsv1";
1677 break;
1678 }
1679 seq_printf(seq, ",jqfmt=%s", fmtname);
1680 }
1681
1682 if (sbi->s_qf_names[USRQUOTA])
1683 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1684
1685 if (sbi->s_qf_names[GRPQUOTA])
1686 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1687
1688 if (test_opt(sb, USRQUOTA))
1689 seq_puts(seq, ",usrquota");
1690
1691 if (test_opt(sb, GRPQUOTA))
1692 seq_puts(seq, ",grpquota");
1693#endif
1694}
1695
5a916be1
TT		 1696static const char *token2str(int token)
1697{
50df9fd5 1698 const struct match_token *t;
5a916be1
TT		 1699
1700 for (t = tokens; t->token != Opt_err; t++)
1701 if (t->token == token && !strchr(t->pattern, '='))
1702 break;
1703 return t->pattern;
1704}
1705
2adf6da8
TT		 1706/*
1707 * Show an option if
1708 * - it's set to a non-default value OR
1709 * - if the per-sb default is different from the global default
1710 */
66acdcf4
TT		 1711static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1712 int nodefs)
2adf6da8 1713{
2adf6da8
TT		 1714 struct ext4_sb_info *sbi = EXT4_SB(sb);
1715 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1716 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1717 const struct mount_opts *m;
66acdcf4 1718 char sep = nodefs ? '\n' : ',';
2adf6da8 1719
66acdcf4
TT		 1720#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1721#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1722
1723 if (sbi->s_sb_block != 1)
5a916be1
TT		 1724 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1725
1726 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1727 int want_set = m->flags & MOPT_SET;
1728 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1729 (m->flags & MOPT_CLEAR_ERR))
1730 continue;
66acdcf4 1731 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1732 continue; /* skip if same as the default */
1733 if ((want_set &&
1734 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1735 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1736 continue; /* select Opt_noFoo vs Opt_Foo */
1737 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1738 }
5a916be1 1739
08cefc7a 1740 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1741 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1742 SEQ_OPTS_PRINT("resuid=%u",
1743 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1744 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1745 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1746 SEQ_OPTS_PRINT("resgid=%u",
1747 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1748 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1749 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1750 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1751 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1752 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1753 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1754 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1755 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1756 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1757 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1758 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1759 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1760 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1761 if (sb->s_flags & MS_I_VERSION)
5a916be1 1762 SEQ_OPTS_PUTS("i_version");
66acdcf4 1763 if (nodefs || sbi->s_stripe)
5a916be1 1764 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1765 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1766 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1767 SEQ_OPTS_PUTS("data=journal");
1768 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1769 SEQ_OPTS_PUTS("data=ordered");
1770 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1771 SEQ_OPTS_PUTS("data=writeback");
1772 }
66acdcf4
TT		 1773 if (nodefs ||
1774 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1775 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1776 sbi->s_inode_readahead_blks);
2adf6da8 1777
66acdcf4
TT		 1778 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1779 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1780 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1781 if (nodefs || sbi->s_max_dir_size_kb)
1782 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1783
1784 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1785 return 0;
1786}
1787
66acdcf4
TT		 1788static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1789{
1790 return _ext4_show_options(seq, root->d_sb, 0);
1791}
1792
1793static int options_seq_show(struct seq_file *seq, void *offset)
1794{
1795 struct super_block *sb = seq->private;
1796 int rc;
1797
1798 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1799 rc = _ext4_show_options(seq, sb, 1);
1800 seq_puts(seq, "\n");
1801 return rc;
1802}
1803
1804static int options_open_fs(struct inode *inode, struct file *file)
1805{
d9dda78b 1806 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1807}
1808
1809static const struct file_operations ext4_seq_options_fops = {
1810 .owner = THIS_MODULE,
1811 .open = options_open_fs,
1812 .read = seq_read,
1813 .llseek = seq_lseek,
1814 .release = single_release,
1815};
1816
617ba13b 1817static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1818 int read_only)
1819{
617ba13b 1820 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1821 int res = 0;
1822
617ba13b 1823 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1824 ext4_msg(sb, KERN_ERR, "revision level too high, "
1825 "forcing read-only mode");
ac27a0ec
DK		 1826 res = MS_RDONLY;
1827 }
1828 if (read_only)
281b5995 1829 goto done;
617ba13b 1830 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1831 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1832 "running e2fsck is recommended");
617ba13b 1833 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1834 ext4_msg(sb, KERN_WARNING,
1835 "warning: mounting fs with errors, "
1836 "running e2fsck is recommended");
ed3ce80a 1837 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1838 le16_to_cpu(es->s_mnt_count) >=
1839 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1840 ext4_msg(sb, KERN_WARNING,
1841 "warning: maximal mount count reached, "
1842 "running e2fsck is recommended");
ac27a0ec
DK		 1843 else if (le32_to_cpu(es->s_checkinterval) &&
1844 (le32_to_cpu(es->s_lastcheck) +
1845 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1846 ext4_msg(sb, KERN_WARNING,
1847 "warning: checktime reached, "
1848 "running e2fsck is recommended");
0b8e58a1 1849 if (!sbi->s_journal)
0390131b 1850 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1851 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1852 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1853 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1854 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1855 ext4_update_dynamic_rev(sb);
0390131b
FM		 1856 if (sbi->s_journal)
1857 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1858
e2d67052 1859 ext4_commit_super(sb, 1);
281b5995 1860done:
ac27a0ec 1861 if (test_opt(sb, DEBUG))
a9df9a49 1862 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1863 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1864 sb->s_blocksize,
1865 sbi->s_groups_count,
617ba13b
MC		 1866 EXT4_BLOCKS_PER_GROUP(sb),
1867 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1868 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1869
7abc52c2 1870 cleancache_init_fs(sb);
ac27a0ec
DK		 1871 return res;
1872}
1873
117fff10
TT		 1874int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1875{
1876 struct ext4_sb_info *sbi = EXT4_SB(sb);
1877 struct flex_groups *new_groups;
1878 int size;
1879
1880 if (!sbi->s_log_groups_per_flex)
1881 return 0;
1882
1883 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1884 if (size <= sbi->s_flex_groups_allocated)
1885 return 0;
1886
1887 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1888 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1889 if (!new_groups) {
1890 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1891 size / (int) sizeof(struct flex_groups));
1892 return -ENOMEM;
1893 }
1894
1895 if (sbi->s_flex_groups) {
1896 memcpy(new_groups, sbi->s_flex_groups,
1897 (sbi->s_flex_groups_allocated *
1898 sizeof(struct flex_groups)));
1899 ext4_kvfree(sbi->s_flex_groups);
1900 }
1901 sbi->s_flex_groups = new_groups;
1902 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1903 return 0;
1904}
1905
772cb7c8
JS		 1906static int ext4_fill_flex_info(struct super_block *sb)
1907{
1908 struct ext4_sb_info *sbi = EXT4_SB(sb);
1909 struct ext4_group_desc *gdp = NULL;
772cb7c8 1910 ext4_group_t flex_group;
d50f2ab6 1911 unsigned int groups_per_flex = 0;
117fff10 1912 int i, err;
772cb7c8 1913
503358ae 1914 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1915 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1916 sbi->s_log_groups_per_flex = 0;
1917 return 1;
1918 }
76495ec1 1919 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1920
117fff10
TT		 1921 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1922 if (err)
9933fc0a 1923 goto failed;
772cb7c8 1924
772cb7c8 1925 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1926 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1927
1928 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1929 atomic_add(ext4_free_inodes_count(sb, gdp),
1930 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1931 atomic64_add(ext4_free_group_clusters(sb, gdp),
1932 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1933 atomic_add(ext4_used_dirs_count(sb, gdp),
1934 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1935 }
1936
1937 return 1;
1938failed:
1939 return 0;
1940}
1941
feb0ab32
DW		 1942static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1943 struct ext4_group_desc *gdp)
717d50e4 1944{
feb0ab32 1945 int offset;
717d50e4 1946 __u16 crc = 0;
feb0ab32 1947 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1948
feb0ab32
DW		 1949 if ((sbi->s_es->s_feature_ro_compat &
1950 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1951 /* Use new metadata_csum algorithm */
d6a77105 1952 __le16 save_csum;
feb0ab32
DW		 1953 __u32 csum32;
1954
d6a77105 1955 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1956 gdp->bg_checksum = 0;
1957 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1958 sizeof(le_group));
1959 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1960 sbi->s_desc_size);
d6a77105 1961 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1962
1963 crc = csum32 & 0xFFFF;
1964 goto out;
717d50e4
AD		 1965 }
1966
feb0ab32
DW		 1967 /* old crc16 code */
1968 offset = offsetof(struct ext4_group_desc, bg_checksum);
1969
1970 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1971 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1972 crc = crc16(crc, (__u8 *)gdp, offset);
1973 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1974 /* for checksum of struct ext4_group_desc do the rest...*/
1975 if ((sbi->s_es->s_feature_incompat &
1976 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1977 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1978 crc = crc16(crc, (__u8 *)gdp + offset,
1979 le16_to_cpu(sbi->s_es->s_desc_size) -
1980 offset);
1981
1982out:
717d50e4
AD		 1983 return cpu_to_le16(crc);
1984}
1985
feb0ab32 1986int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 1987 struct ext4_group_desc *gdp)
1988{
feb0ab32
DW		 1989 if (ext4_has_group_desc_csum(sb) &&
1990 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1991 block_group, gdp)))
717d50e4
AD		 1992 return 0;
1993
1994 return 1;
1995}
1996
feb0ab32
DW		 1997void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1998 struct ext4_group_desc *gdp)
1999{
2000 if (!ext4_has_group_desc_csum(sb))
2001 return;
2002 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2003}
2004
ac27a0ec 2005/* Called at mount-time, super-block is locked */
bfff6873
LC		 2006static int ext4_check_descriptors(struct super_block *sb,
2007 ext4_group_t *first_not_zeroed)
ac27a0ec 2008{
617ba13b
MC		 2009 struct ext4_sb_info *sbi = EXT4_SB(sb);
2010 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2011 ext4_fsblk_t last_block;
bd81d8ee
LV		 2012 ext4_fsblk_t block_bitmap;
2013 ext4_fsblk_t inode_bitmap;
2014 ext4_fsblk_t inode_table;
ce421581 2015 int flexbg_flag = 0;
bfff6873 2016 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2017
ce421581
JS		 2018 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2019 flexbg_flag = 1;
2020
af5bc92d 2021 ext4_debug("Checking group descriptors");
ac27a0ec 2022
197cd65a
AM		 2023 for (i = 0; i < sbi->s_groups_count; i++) {
2024 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2025
ce421581 2026 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2027 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2028 else
2029 last_block = first_block +
617ba13b 2030 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2031
bfff6873
LC		 2032 if ((grp == sbi->s_groups_count) &&
2033 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2034 grp = i;
2035
8fadc143 2036 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2037 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2038 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2039 "Block bitmap for group %u not in group "
b31e1552 2040 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2041 return 0;
2042 }
8fadc143 2043 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2044 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2045 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2046 "Inode bitmap for group %u not in group "
b31e1552 2047 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2048 return 0;
2049 }
8fadc143 2050 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2051 if (inode_table < first_block ||
2b2d6d01 2052 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2053 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2054 "Inode table for group %u not in group "
b31e1552 2055 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2056 return 0;
2057 }
955ce5f5 2058 ext4_lock_group(sb, i);
feb0ab32 2059 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2060 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2061 "Checksum for group %u failed (%u!=%u)",
2062 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2063 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2064 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2065 ext4_unlock_group(sb, i);
8a266467 2066 return 0;
7ee1ec4c 2067 }
717d50e4 2068 }
955ce5f5 2069 ext4_unlock_group(sb, i);
ce421581
JS		 2070 if (!flexbg_flag)
2071 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2072 }
bfff6873
LC		 2073 if (NULL != first_not_zeroed)
2074 *first_not_zeroed = grp;
ac27a0ec 2075
5dee5437
TT		 2076 ext4_free_blocks_count_set(sbi->s_es,
2077 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2078 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2079 return 1;
2080}
2081
617ba13b 2082/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2083 * the superblock) which were deleted from all directories, but held open by
2084 * a process at the time of a crash. We walk the list and try to delete these
2085 * inodes at recovery time (only with a read-write filesystem).
2086 *
2087 * In order to keep the orphan inode chain consistent during traversal (in
2088 * case of crash during recovery), we link each inode into the superblock
2089 * orphan list_head and handle it the same way as an inode deletion during
2090 * normal operation (which journals the operations for us).
2091 *
2092 * We only do an iget() and an iput() on each inode, which is very safe if we
2093 * accidentally point at an in-use or already deleted inode. The worst that
2094 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2095 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2096 * e2fsck was run on this filesystem, and it must have already done the orphan
2097 * inode cleanup for us, so we can safely abort without any further action.
2098 */
2b2d6d01
TT		 2099static void ext4_orphan_cleanup(struct super_block *sb,
2100 struct ext4_super_block *es)
ac27a0ec
DK		 2101{
2102 unsigned int s_flags = sb->s_flags;
2103 int nr_orphans = 0, nr_truncates = 0;
2104#ifdef CONFIG_QUOTA
2105 int i;
2106#endif
2107 if (!es->s_last_orphan) {
2108 jbd_debug(4, "no orphan inodes to clean up\n");
2109 return;
2110 }
2111
a8f48a95 2112 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2113 ext4_msg(sb, KERN_ERR, "write access "
2114 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2115 return;
2116 }
2117
d39195c3
AG		 2118 /* Check if feature set would not allow a r/w mount */
2119 if (!ext4_feature_set_ok(sb, 0)) {
2120 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2121 "unknown ROCOMPAT features");
2122 return;
2123 }
2124
617ba13b 2125 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2126 /* don't clear list on RO mount w/ errors */
2127 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2128 jbd_debug(1, "Errors on filesystem, "
2129 "clearing orphan list.\n");
c25f9bc6
ES		 2130 es->s_last_orphan = 0;
2131 }
ac27a0ec
DK		 2132 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2133 return;
2134 }
2135
2136 if (s_flags & MS_RDONLY) {
b31e1552 2137 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2138 sb->s_flags &= ~MS_RDONLY;
2139 }
2140#ifdef CONFIG_QUOTA
2141 /* Needed for iput() to work correctly and not trash data */
2142 sb->s_flags |= MS_ACTIVE;
2143 /* Turn on quotas so that they are updated correctly */
2144 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2145 if (EXT4_SB(sb)->s_qf_names[i]) {
2146 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2147 if (ret < 0)
b31e1552
ES		 2148 ext4_msg(sb, KERN_ERR,
2149 "Cannot turn on journaled "
2150 "quota: error %d", ret);
ac27a0ec
DK		 2151 }
2152 }
2153#endif
2154
2155 while (es->s_last_orphan) {
2156 struct inode *inode;
2157
97bd42b9
JB		 2158 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2159 if (IS_ERR(inode)) {
ac27a0ec
DK		 2160 es->s_last_orphan = 0;
2161 break;
2162 }
2163
617ba13b 2164 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2165 dquot_initialize(inode);
ac27a0ec 2166 if (inode->i_nlink) {
b31e1552
ES		 2167 ext4_msg(sb, KERN_DEBUG,
2168 "%s: truncating inode %lu to %lld bytes",
46e665e9 2169 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2170 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2171 inode->i_ino, inode->i_size);
721e3eba 2172 mutex_lock(&inode->i_mutex);
55f252c9 2173 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2174 ext4_truncate(inode);
721e3eba 2175 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2176 nr_truncates++;
2177 } else {
b31e1552
ES		 2178 ext4_msg(sb, KERN_DEBUG,
2179 "%s: deleting unreferenced inode %lu",
46e665e9 2180 __func__, inode->i_ino);
ac27a0ec
DK		 2181 jbd_debug(2, "deleting unreferenced inode %lu\n",
2182 inode->i_ino);
2183 nr_orphans++;
2184 }
2185 iput(inode); /* The delete magic happens here! */
2186 }
2187
2b2d6d01 2188#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2189
2190 if (nr_orphans)
b31e1552
ES		 2191 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2192 PLURAL(nr_orphans));
ac27a0ec 2193 if (nr_truncates)
b31e1552
ES		 2194 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2195 PLURAL(nr_truncates));
ac27a0ec
DK		 2196#ifdef CONFIG_QUOTA
2197 /* Turn quotas off */
2198 for (i = 0; i < MAXQUOTAS; i++) {
2199 if (sb_dqopt(sb)->files[i])
287a8095 2200 dquot_quota_off(sb, i);
ac27a0ec
DK		 2201 }
2202#endif
2203 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2204}
0b8e58a1 2205
cd2291a4
ES		 2206/*
2207 * Maximal extent format file size.
2208 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2209 * extent format containers, within a sector_t, and within i_blocks
2210 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2211 * so that won't be a limiting factor.
2212 *
f17722f9
LC		 2213 * However there is other limiting factor. We do store extents in the form
2214 * of starting block and length, hence the resulting length of the extent
2215 * covering maximum file size must fit into on-disk format containers as
2216 * well. Given that length is always by 1 unit bigger than max unit (because
2217 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2218 *
cd2291a4
ES		 2219 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2220 */
f287a1a5 2221static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2222{
2223 loff_t res;
2224 loff_t upper_limit = MAX_LFS_FILESIZE;
2225
2226 /* small i_blocks in vfs inode? */
f287a1a5 2227 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2228 /*
90c699a9 2229 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2230 * i_block represent total blocks in 512 bytes
2231 * 32 == size of vfs inode i_blocks * 8
2232 */
2233 upper_limit = (1LL << 32) - 1;
2234
2235 /* total blocks in file system block size */
2236 upper_limit >>= (blkbits - 9);
2237 upper_limit <<= blkbits;
2238 }
2239
f17722f9
LC		 2240 /*
2241 * 32-bit extent-start container, ee_block. We lower the maxbytes
2242 * by one fs block, so ee_len can cover the extent of maximum file
2243 * size
2244 */
2245 res = (1LL << 32) - 1;
cd2291a4 2246 res <<= blkbits;
cd2291a4
ES		 2247
2248 /* Sanity check against vm- & vfs- imposed limits */
2249 if (res > upper_limit)
2250 res = upper_limit;
2251
2252 return res;
2253}
ac27a0ec 2254
ac27a0ec 2255/*
cd2291a4 2256 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2257 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2258 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2259 */
f287a1a5 2260static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2261{
617ba13b 2262 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2263 int meta_blocks;
2264 loff_t upper_limit;
0b8e58a1
AD		 2265 /* This is calculated to be the largest file size for a dense, block
2266 * mapped file such that the file's total number of 512-byte sectors,
2267 * including data and all indirect blocks, does not exceed (2^48 - 1).
2268 *
2269 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2270 * number of 512-byte sectors of the file.
0fc1b451
AK		 2271 */
2272
f287a1a5 2273 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2274 /*
90c699a9 2275 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2276 * the inode i_block field represents total file blocks in
2277 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2278 */
2279 upper_limit = (1LL << 32) - 1;
2280
2281 /* total blocks in file system block size */
2282 upper_limit >>= (bits - 9);
2283
2284 } else {
8180a562
AK		 2285 /*
2286 * We use 48 bit ext4_inode i_blocks
2287 * With EXT4_HUGE_FILE_FL set the i_blocks
2288 * represent total number of blocks in
2289 * file system block size
2290 */
0fc1b451
AK		 2291 upper_limit = (1LL << 48) - 1;
2292
0fc1b451
AK		 2293 }
2294
2295 /* indirect blocks */
2296 meta_blocks = 1;
2297 /* double indirect blocks */
2298 meta_blocks += 1 + (1LL << (bits-2));
2299 /* tripple indirect blocks */
2300 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2301
2302 upper_limit -= meta_blocks;
2303 upper_limit <<= bits;
ac27a0ec
DK		 2304
2305 res += 1LL << (bits-2);
2306 res += 1LL << (2*(bits-2));
2307 res += 1LL << (3*(bits-2));
2308 res <<= bits;
2309 if (res > upper_limit)
2310 res = upper_limit;
0fc1b451
AK		 2311
2312 if (res > MAX_LFS_FILESIZE)
2313 res = MAX_LFS_FILESIZE;
2314
ac27a0ec
DK		 2315 return res;
2316}
2317
617ba13b 2318static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2319 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2320{
617ba13b 2321 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2322 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2323 int has_super = 0;
2324
2325 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2326
617ba13b 2327 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2328 nr < first_meta_bg)
70bbb3e0 2329 return logical_sb_block + nr + 1;
ac27a0ec 2330 bg = sbi->s_desc_per_block * nr;
617ba13b 2331 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2332 has_super = 1;
0b8e58a1 2333
617ba13b 2334 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2335}
2336
c9de560d
AT		 2337/**
2338 * ext4_get_stripe_size: Get the stripe size.
2339 * @sbi: In memory super block info
2340 *
2341 * If we have specified it via mount option, then
2342 * use the mount option value. If the value specified at mount time is
2343 * greater than the blocks per group use the super block value.
2344 * If the super block value is greater than blocks per group return 0.
2345 * Allocator needs it be less than blocks per group.
2346 *
2347 */
2348static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2349{
2350 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2351 unsigned long stripe_width =
2352 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2353 int ret;
c9de560d
AT		 2354
2355 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2356 ret = sbi->s_stripe;
2357 else if (stripe_width <= sbi->s_blocks_per_group)
2358 ret = stripe_width;
2359 else if (stride <= sbi->s_blocks_per_group)
2360 ret = stride;
2361 else
2362 ret = 0;
c9de560d 2363
3eb08658
DE		 2364 /*
2365 * If the stripe width is 1, this makes no sense and
2366 * we set it to 0 to turn off stripe handling code.
2367 */
2368 if (ret <= 1)
2369 ret = 0;
c9de560d 2370
3eb08658 2371 return ret;
c9de560d 2372}
ac27a0ec 2373
3197ebdb
TT		 2374/* sysfs supprt */
2375
2376struct ext4_attr {
2377 struct attribute attr;
2378 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2379 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2380 const char *, size_t);
2381 int offset;
2382};
2383
27dd4385
LC		 2384static int parse_strtoull(const char *buf,
2385 unsigned long long max, unsigned long long *value)
2386{
2387 int ret;
2388
2389 ret = kstrtoull(skip_spaces(buf), 0, value);
2390 if (!ret && *value > max)
2391 ret = -EINVAL;
2392 return ret;
2393}
2394
3197ebdb
TT		 2395static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2396 struct ext4_sb_info *sbi,
2397 char *buf)
2398{
2399 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2400 (s64) EXT4_C2B(sbi,
2401 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2402}
2403
2404static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2405 struct ext4_sb_info *sbi, char *buf)
2406{
2407 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2408
f613dfcb
TT		 2409 if (!sb->s_bdev->bd_part)
2410 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2411 return snprintf(buf, PAGE_SIZE, "%lu\n",
2412 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2413 sbi->s_sectors_written_start) >> 1);
2414}
2415
2416static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2417 struct ext4_sb_info *sbi, char *buf)
2418{
2419 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2420
f613dfcb
TT		 2421 if (!sb->s_bdev->bd_part)
2422 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2423 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2424 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2425 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2426 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2427}
2428
2429static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2430 struct ext4_sb_info *sbi,
2431 const char *buf, size_t count)
2432{
2433 unsigned long t;
e1091b15 2434 int ret;
3197ebdb 2435
e1091b15
LC		 2436 ret = kstrtoul(skip_spaces(buf), 0, &t);
2437 if (ret)
2438 return ret;
3197ebdb 2439
e1091b15 2440 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2441 return -EINVAL;
2442
2443 sbi->s_inode_readahead_blks = t;
2444 return count;
2445}
2446
2447static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2448 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2449{
2450 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2451
2452 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2453}
2454
2455static ssize_t sbi_ui_store(struct ext4_attr *a,
2456 struct ext4_sb_info *sbi,
2457 const char *buf, size_t count)
2458{
2459 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2460 unsigned long t;
e1091b15 2461 int ret;
3197ebdb 2462
e1091b15
LC		 2463 ret = kstrtoul(skip_spaces(buf), 0, &t);
2464 if (ret)
2465 return ret;
3197ebdb
TT		 2466 *ui = t;
2467 return count;
2468}
2469
27dd4385
LC		 2470static ssize_t reserved_clusters_show(struct ext4_attr *a,
2471 struct ext4_sb_info *sbi, char *buf)
2472{
2473 return snprintf(buf, PAGE_SIZE, "%llu\n",
2474 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2475}
2476
2477static ssize_t reserved_clusters_store(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi,
2479 const char *buf, size_t count)
2480{
2481 unsigned long long val;
2482 int ret;
2483
2484 if (parse_strtoull(buf, -1ULL, &val))
2485 return -EINVAL;
2486 ret = ext4_reserve_clusters(sbi, val);
2487
2488 return ret ? ret : count;
2489}
2490
2c0544b2
TT		 2491static ssize_t trigger_test_error(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi,
2493 const char *buf, size_t count)
2494{
2495 int len = count;
2496
2497 if (!capable(CAP_SYS_ADMIN))
2498 return -EPERM;
2499
2500 if (len && buf[len-1] == '\n')
2501 len--;
2502
2503 if (len)
2504 ext4_error(sbi->s_sb, "%.*s", len, buf);
2505 return count;
2506}
2507
3197ebdb
TT		 2508#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2509static struct ext4_attr ext4_attr_##_name = { \
2510 .attr = {.name = __stringify(_name), .mode = _mode }, \
2511 .show = _show, \
2512 .store = _store, \
2513 .offset = offsetof(struct ext4_sb_info, _elname), \
2514}
2515#define EXT4_ATTR(name, mode, show, store) \
2516static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2517
857ac889 2518#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2519#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2520#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2521#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2522 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2523#define ATTR_LIST(name) &ext4_attr_##name.attr
2524
2525EXT4_RO_ATTR(delayed_allocation_blocks);
2526EXT4_RO_ATTR(session_write_kbytes);
2527EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2528EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2529EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2530 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2531EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2532EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2533EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2534EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2535EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2536EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2537EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2538EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2539EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2540EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2541
2542static struct attribute *ext4_attrs[] = {
2543 ATTR_LIST(delayed_allocation_blocks),
2544 ATTR_LIST(session_write_kbytes),
2545 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2546 ATTR_LIST(reserved_clusters),
3197ebdb 2547 ATTR_LIST(inode_readahead_blks),
11013911 2548 ATTR_LIST(inode_goal),
3197ebdb
TT		 2549 ATTR_LIST(mb_stats),
2550 ATTR_LIST(mb_max_to_scan),
2551 ATTR_LIST(mb_min_to_scan),
2552 ATTR_LIST(mb_order2_req),
2553 ATTR_LIST(mb_stream_req),
2554 ATTR_LIST(mb_group_prealloc),
55138e0b 2555 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2556 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2557 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2558 NULL,
2559};
2560
857ac889
LC		 2561/* Features this copy of ext4 supports */
2562EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2563EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2564EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2565
2566static struct attribute *ext4_feat_attrs[] = {
2567 ATTR_LIST(lazy_itable_init),
27ee40df 2568 ATTR_LIST(batched_discard),
5e7bbef1 2569 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2570 NULL,
2571};
2572
3197ebdb
TT		 2573static ssize_t ext4_attr_show(struct kobject *kobj,
2574 struct attribute *attr, char *buf)
2575{
2576 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2577 s_kobj);
2578 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2579
2580 return a->show ? a->show(a, sbi, buf) : 0;
2581}
2582
2583static ssize_t ext4_attr_store(struct kobject *kobj,
2584 struct attribute *attr,
2585 const char *buf, size_t len)
2586{
2587 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2588 s_kobj);
2589 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2590
2591 return a->store ? a->store(a, sbi, buf, len) : 0;
2592}
2593
2594static void ext4_sb_release(struct kobject *kobj)
2595{
2596 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2597 s_kobj);
2598 complete(&sbi->s_kobj_unregister);
2599}
2600
52cf25d0 2601static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2602 .show = ext4_attr_show,
2603 .store = ext4_attr_store,
2604};
2605
2606static struct kobj_type ext4_ktype = {
2607 .default_attrs = ext4_attrs,
2608 .sysfs_ops = &ext4_attr_ops,
2609 .release = ext4_sb_release,
2610};
2611
857ac889
LC		 2612static void ext4_feat_release(struct kobject *kobj)
2613{
2614 complete(&ext4_feat->f_kobj_unregister);
2615}
2616
2617static struct kobj_type ext4_feat_ktype = {
2618 .default_attrs = ext4_feat_attrs,
2619 .sysfs_ops = &ext4_attr_ops,
2620 .release = ext4_feat_release,
2621};
2622
a13fb1a4
ES		 2623/*
2624 * Check whether this filesystem can be mounted based on
2625 * the features present and the RDONLY/RDWR mount requested.
2626 * Returns 1 if this filesystem can be mounted as requested,
2627 * 0 if it cannot be.
2628 */
2629static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2630{
2631 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2632 ext4_msg(sb, KERN_ERR,
2633 "Couldn't mount because of "
2634 "unsupported optional features (%x)",
2635 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2636 ~EXT4_FEATURE_INCOMPAT_SUPP));
2637 return 0;
2638 }
2639
2640 if (readonly)
2641 return 1;
2642
2643 /* Check that feature set is OK for a read-write mount */
2644 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2645 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2646 "unsupported optional features (%x)",
2647 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2648 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2649 return 0;
2650 }
2651 /*
2652 * Large file size enabled file system can only be mounted
2653 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2654 */
2655 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2656 if (sizeof(blkcnt_t) < sizeof(u64)) {
2657 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2658 "cannot be mounted RDWR without "
2659 "CONFIG_LBDAF");
2660 return 0;
2661 }
2662 }
bab08ab9
TT		 2663 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2664 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2665 ext4_msg(sb, KERN_ERR,
2666 "Can't support bigalloc feature without "
2667 "extents feature\n");
2668 return 0;
2669 }
7c319d32
AK		 2670
2671#ifndef CONFIG_QUOTA
2672 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2673 !readonly) {
2674 ext4_msg(sb, KERN_ERR,
2675 "Filesystem with quota feature cannot be mounted RDWR "
2676 "without CONFIG_QUOTA");
2677 return 0;
2678 }
2679#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2680 return 1;
2681}
2682
66e61a9e
TT		 2683/*
2684 * This function is called once a day if we have errors logged
2685 * on the file system
2686 */
2687static void print_daily_error_info(unsigned long arg)
2688{
2689 struct super_block *sb = (struct super_block *) arg;
2690 struct ext4_sb_info *sbi;
2691 struct ext4_super_block *es;
2692
2693 sbi = EXT4_SB(sb);
2694 es = sbi->s_es;
2695
2696 if (es->s_error_count)
2697 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2698 le32_to_cpu(es->s_error_count));
2699 if (es->s_first_error_time) {
2700 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2701 sb->s_id, le32_to_cpu(es->s_first_error_time),
2702 (int) sizeof(es->s_first_error_func),
2703 es->s_first_error_func,
2704 le32_to_cpu(es->s_first_error_line));
2705 if (es->s_first_error_ino)
2706 printk(": inode %u",
2707 le32_to_cpu(es->s_first_error_ino));
2708 if (es->s_first_error_block)
2709 printk(": block %llu", (unsigned long long)
2710 le64_to_cpu(es->s_first_error_block));
2711 printk("\n");
2712 }
2713 if (es->s_last_error_time) {
2714 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2715 sb->s_id, le32_to_cpu(es->s_last_error_time),
2716 (int) sizeof(es->s_last_error_func),
2717 es->s_last_error_func,
2718 le32_to_cpu(es->s_last_error_line));
2719 if (es->s_last_error_ino)
2720 printk(": inode %u",
2721 le32_to_cpu(es->s_last_error_ino));
2722 if (es->s_last_error_block)
2723 printk(": block %llu", (unsigned long long)
2724 le64_to_cpu(es->s_last_error_block));
2725 printk("\n");
2726 }
2727 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2728}
2729
bfff6873
LC		 2730/* Find next suitable group and run ext4_init_inode_table */
2731static int ext4_run_li_request(struct ext4_li_request *elr)
2732{
2733 struct ext4_group_desc *gdp = NULL;
2734 ext4_group_t group, ngroups;
2735 struct super_block *sb;
2736 unsigned long timeout = 0;
2737 int ret = 0;
2738
2739 sb = elr->lr_super;
2740 ngroups = EXT4_SB(sb)->s_groups_count;
2741
8e8ad8a5 2742 sb_start_write(sb);
bfff6873
LC		 2743 for (group = elr->lr_next_group; group < ngroups; group++) {
2744 gdp = ext4_get_group_desc(sb, group, NULL);
2745 if (!gdp) {
2746 ret = 1;
2747 break;
2748 }
2749
2750 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2751 break;
2752 }
2753
7f511862 2754 if (group >= ngroups)
bfff6873
LC		 2755 ret = 1;
2756
2757 if (!ret) {
2758 timeout = jiffies;
2759 ret = ext4_init_inode_table(sb, group,
2760 elr->lr_timeout ? 0 : 1);
2761 if (elr->lr_timeout == 0) {
51ce6511
LC		 2762 timeout = (jiffies - timeout) *
2763 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2764 elr->lr_timeout = timeout;
2765 }
2766 elr->lr_next_sched = jiffies + elr->lr_timeout;
2767 elr->lr_next_group = group + 1;
2768 }
8e8ad8a5 2769 sb_end_write(sb);
bfff6873
LC		 2770
2771 return ret;
2772}
2773
2774/*
2775 * Remove lr_request from the list_request and free the
4ed5c033 2776 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2777 */
2778static void ext4_remove_li_request(struct ext4_li_request *elr)
2779{
2780 struct ext4_sb_info *sbi;
2781
2782 if (!elr)
2783 return;
2784
2785 sbi = elr->lr_sbi;
2786
2787 list_del(&elr->lr_request);
2788 sbi->s_li_request = NULL;
2789 kfree(elr);
2790}
2791
2792static void ext4_unregister_li_request(struct super_block *sb)
2793{
1bb933fb
LC		 2794 mutex_lock(&ext4_li_mtx);
2795 if (!ext4_li_info) {
2796 mutex_unlock(&ext4_li_mtx);
bfff6873 2797 return;
1bb933fb 2798 }
bfff6873
LC		 2799
2800 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2801 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2802 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2803 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2804}
2805
8f1f7453
ES		 2806static struct task_struct *ext4_lazyinit_task;
2807
bfff6873
LC		 2808/*
2809 * This is the function where ext4lazyinit thread lives. It walks
2810 * through the request list searching for next scheduled filesystem.
2811 * When such a fs is found, run the lazy initialization request
2812 * (ext4_rn_li_request) and keep track of the time spend in this
2813 * function. Based on that time we compute next schedule time of
2814 * the request. When walking through the list is complete, compute
2815 * next waking time and put itself into sleep.
2816 */
2817static int ext4_lazyinit_thread(void *arg)
2818{
2819 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2820 struct list_head *pos, *n;
2821 struct ext4_li_request *elr;
4ed5c033 2822 unsigned long next_wakeup, cur;
bfff6873
LC		 2823
2824 BUG_ON(NULL == eli);
2825
bfff6873
LC		 2826cont_thread:
2827 while (true) {
2828 next_wakeup = MAX_JIFFY_OFFSET;
2829
2830 mutex_lock(&eli->li_list_mtx);
2831 if (list_empty(&eli->li_request_list)) {
2832 mutex_unlock(&eli->li_list_mtx);
2833 goto exit_thread;
2834 }
2835
2836 list_for_each_safe(pos, n, &eli->li_request_list) {
2837 elr = list_entry(pos, struct ext4_li_request,
2838 lr_request);
2839
b2c78cd0
TT		 2840 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2841 if (ext4_run_li_request(elr) != 0) {
2842 /* error, remove the lazy_init job */
2843 ext4_remove_li_request(elr);
2844 continue;
2845 }
bfff6873
LC		 2846 }
2847
2848 if (time_before(elr->lr_next_sched, next_wakeup))
2849 next_wakeup = elr->lr_next_sched;
2850 }
2851 mutex_unlock(&eli->li_list_mtx);
2852
a0acae0e 2853 try_to_freeze();
bfff6873 2854
4ed5c033
LC		 2855 cur = jiffies;
2856 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2857 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2858 cond_resched();
2859 continue;
2860 }
2861
4ed5c033
LC		 2862 schedule_timeout_interruptible(next_wakeup - cur);
2863
8f1f7453
ES		 2864 if (kthread_should_stop()) {
2865 ext4_clear_request_list();
2866 goto exit_thread;
2867 }
bfff6873
LC		 2868 }
2869
2870exit_thread:
2871 /*
2872 * It looks like the request list is empty, but we need
2873 * to check it under the li_list_mtx lock, to prevent any
2874 * additions into it, and of course we should lock ext4_li_mtx
2875 * to atomically free the list and ext4_li_info, because at
2876 * this point another ext4 filesystem could be registering
2877 * new one.
2878 */
2879 mutex_lock(&ext4_li_mtx);
2880 mutex_lock(&eli->li_list_mtx);
2881 if (!list_empty(&eli->li_request_list)) {
2882 mutex_unlock(&eli->li_list_mtx);
2883 mutex_unlock(&ext4_li_mtx);
2884 goto cont_thread;
2885 }
2886 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2887 kfree(ext4_li_info);
2888 ext4_li_info = NULL;
2889 mutex_unlock(&ext4_li_mtx);
2890
2891 return 0;
2892}
2893
2894static void ext4_clear_request_list(void)
2895{
2896 struct list_head *pos, *n;
2897 struct ext4_li_request *elr;
2898
2899 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2900 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2901 elr = list_entry(pos, struct ext4_li_request,
2902 lr_request);
2903 ext4_remove_li_request(elr);
2904 }
2905 mutex_unlock(&ext4_li_info->li_list_mtx);
2906}
2907
2908static int ext4_run_lazyinit_thread(void)
2909{
8f1f7453
ES		 2910 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2911 ext4_li_info, "ext4lazyinit");
2912 if (IS_ERR(ext4_lazyinit_task)) {
2913 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2914 ext4_clear_request_list();
bfff6873
LC		 2915 kfree(ext4_li_info);
2916 ext4_li_info = NULL;
92b97816 2917 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2918 "initialization thread\n",
2919 err);
2920 return err;
2921 }
2922 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2923 return 0;
2924}
2925
2926/*
2927 * Check whether it make sense to run itable init. thread or not.
2928 * If there is at least one uninitialized inode table, return
2929 * corresponding group number, else the loop goes through all
2930 * groups and return total number of groups.
2931 */
2932static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2933{
2934 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2935 struct ext4_group_desc *gdp = NULL;
2936
2937 for (group = 0; group < ngroups; group++) {
2938 gdp = ext4_get_group_desc(sb, group, NULL);
2939 if (!gdp)
2940 continue;
2941
2942 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2943 break;
2944 }
2945
2946 return group;
2947}
2948
2949static int ext4_li_info_new(void)
2950{
2951 struct ext4_lazy_init *eli = NULL;
2952
2953 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2954 if (!eli)
2955 return -ENOMEM;
2956
bfff6873
LC		 2957 INIT_LIST_HEAD(&eli->li_request_list);
2958 mutex_init(&eli->li_list_mtx);
2959
bfff6873
LC		 2960 eli->li_state |= EXT4_LAZYINIT_QUIT;
2961
2962 ext4_li_info = eli;
2963
2964 return 0;
2965}
2966
2967static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2968 ext4_group_t start)
2969{
2970 struct ext4_sb_info *sbi = EXT4_SB(sb);
2971 struct ext4_li_request *elr;
2972 unsigned long rnd;
2973
2974 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2975 if (!elr)
2976 return NULL;
2977
2978 elr->lr_super = sb;
2979 elr->lr_sbi = sbi;
2980 elr->lr_next_group = start;
2981
2982 /*
2983 * Randomize first schedule time of the request to
2984 * spread the inode table initialization requests
2985 * better.
2986 */
2987 get_random_bytes(&rnd, sizeof(rnd));
2988 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2989 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2990
2991 return elr;
2992}
2993
7f511862
TT		 2994int ext4_register_li_request(struct super_block *sb,
2995 ext4_group_t first_not_zeroed)
bfff6873
LC		 2996{
2997 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2998 struct ext4_li_request *elr = NULL;
bfff6873 2999 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3000 int ret = 0;
bfff6873 3001
7f511862 3002 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3003 if (sbi->s_li_request != NULL) {
3004 /*
3005 * Reset timeout so it can be computed again, because
3006 * s_li_wait_mult might have changed.
3007 */
3008 sbi->s_li_request->lr_timeout = 0;
7f511862 3009 goto out;
51ce6511 3010 }
bfff6873
LC		 3011
3012 if (first_not_zeroed == ngroups ||
3013 (sb->s_flags & MS_RDONLY) ||
55ff3840 3014 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3015 goto out;
bfff6873
LC		 3016
3017 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3018 if (!elr) {
3019 ret = -ENOMEM;
3020 goto out;
3021 }
bfff6873
LC		 3022
3023 if (NULL == ext4_li_info) {
3024 ret = ext4_li_info_new();
3025 if (ret)
3026 goto out;
3027 }
3028
3029 mutex_lock(&ext4_li_info->li_list_mtx);
3030 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3031 mutex_unlock(&ext4_li_info->li_list_mtx);
3032
3033 sbi->s_li_request = elr;
46e4690b
TM		 3034 /*
3035 * set elr to NULL here since it has been inserted to
3036 * the request_list and the removal and free of it is
3037 * handled by ext4_clear_request_list from now on.
3038 */
3039 elr = NULL;
bfff6873
LC		 3040
3041 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3042 ret = ext4_run_lazyinit_thread();
3043 if (ret)
3044 goto out;
3045 }
bfff6873 3046out:
beed5ecb
NK		 3047 mutex_unlock(&ext4_li_mtx);
3048 if (ret)
bfff6873 3049 kfree(elr);
bfff6873
LC		 3050 return ret;
3051}
3052
3053/*
3054 * We do not need to lock anything since this is called on
3055 * module unload.
3056 */
3057static void ext4_destroy_lazyinit_thread(void)
3058{
3059 /*
3060 * If thread exited earlier
3061 * there's nothing to be done.
3062 */
8f1f7453 3063 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3064 return;
3065
8f1f7453 3066 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3067}
3068
25ed6e8a
DW		 3069static int set_journal_csum_feature_set(struct super_block *sb)
3070{
3071 int ret = 1;
3072 int compat, incompat;
3073 struct ext4_sb_info *sbi = EXT4_SB(sb);
3074
3075 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3076 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3077 /* journal checksum v2 */
3078 compat = 0;
3079 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3080 } else {
3081 /* journal checksum v1 */
3082 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3083 incompat = 0;
3084 }
3085
3086 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3087 ret = jbd2_journal_set_features(sbi->s_journal,
3088 compat, 0,
3089 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3090 incompat);
3091 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3092 ret = jbd2_journal_set_features(sbi->s_journal,
3093 compat, 0,
3094 incompat);
3095 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3096 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3097 } else {
3098 jbd2_journal_clear_features(sbi->s_journal,
3099 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3100 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3101 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3102 }
3103
3104 return ret;
3105}
3106
952fc18e
TT		 3107/*
3108 * Note: calculating the overhead so we can be compatible with
3109 * historical BSD practice is quite difficult in the face of
3110 * clusters/bigalloc. This is because multiple metadata blocks from
3111 * different block group can end up in the same allocation cluster.
3112 * Calculating the exact overhead in the face of clustered allocation
3113 * requires either O(all block bitmaps) in memory or O(number of block
3114 * groups**2) in time. We will still calculate the superblock for
3115 * older file systems --- and if we come across with a bigalloc file
3116 * system with zero in s_overhead_clusters the estimate will be close to
3117 * correct especially for very large cluster sizes --- but for newer
3118 * file systems, it's better to calculate this figure once at mkfs
3119 * time, and store it in the superblock. If the superblock value is
3120 * present (even for non-bigalloc file systems), we will use it.
3121 */
3122static int count_overhead(struct super_block *sb, ext4_group_t grp,
3123 char *buf)
3124{
3125 struct ext4_sb_info *sbi = EXT4_SB(sb);
3126 struct ext4_group_desc *gdp;
3127 ext4_fsblk_t first_block, last_block, b;
3128 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3129 int s, j, count = 0;
3130
0548bbb8
TT		 3131 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3132 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3133 sbi->s_itb_per_group + 2);
3134
952fc18e
TT		 3135 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3136 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3137 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3138 for (i = 0; i < ngroups; i++) {
3139 gdp = ext4_get_group_desc(sb, i, NULL);
3140 b = ext4_block_bitmap(sb, gdp);
3141 if (b >= first_block && b <= last_block) {
3142 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3143 count++;
3144 }
3145 b = ext4_inode_bitmap(sb, gdp);
3146 if (b >= first_block && b <= last_block) {
3147 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3148 count++;
3149 }
3150 b = ext4_inode_table(sb, gdp);
3151 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3152 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3153 int c = EXT4_B2C(sbi, b - first_block);
3154 ext4_set_bit(c, buf);
3155 count++;
3156 }
3157 if (i != grp)
3158 continue;
3159 s = 0;
3160 if (ext4_bg_has_super(sb, grp)) {
3161 ext4_set_bit(s++, buf);
3162 count++;
3163 }
3164 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3165 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3166 count++;
3167 }
3168 }
3169 if (!count)
3170 return 0;
3171 return EXT4_CLUSTERS_PER_GROUP(sb) -
3172 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3173}
3174
3175/*
3176 * Compute the overhead and stash it in sbi->s_overhead
3177 */
3178int ext4_calculate_overhead(struct super_block *sb)
3179{
3180 struct ext4_sb_info *sbi = EXT4_SB(sb);
3181 struct ext4_super_block *es = sbi->s_es;
3182 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3183 ext4_fsblk_t overhead = 0;
3184 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3185
952fc18e
TT		 3186 if (!buf)
3187 return -ENOMEM;
3188
3189 /*
3190 * Compute the overhead (FS structures). This is constant
3191 * for a given filesystem unless the number of block groups
3192 * changes so we cache the previous value until it does.
3193 */
3194
3195 /*
3196 * All of the blocks before first_data_block are overhead
3197 */
3198 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3199
3200 /*
3201 * Add the overhead found in each block group
3202 */
3203 for (i = 0; i < ngroups; i++) {
3204 int blks;
3205
3206 blks = count_overhead(sb, i, buf);
3207 overhead += blks;
3208 if (blks)
3209 memset(buf, 0, PAGE_SIZE);
3210 cond_resched();
3211 }
0875a2b4
ES		 3212 /* Add the journal blocks as well */
3213 if (sbi->s_journal)
810da240 3214 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3215
952fc18e
TT		 3216 sbi->s_overhead = overhead;
3217 smp_wmb();
3218 free_page((unsigned long) buf);
3219 return 0;
3220}
3221
27dd4385
LC		 3222
3223static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3224{
3225 ext4_fsblk_t resv_clusters;
3226
3227 /*
3228 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3229 * This should cover the situations where we can not afford to run
3230 * out of space like for example punch hole, or converting
3231 * uninitialized extents in delalloc path. In most cases such
3232 * allocation would require 1, or 2 blocks, higher numbers are
3233 * very rare.
3234 */
3235 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3236
3237 do_div(resv_clusters, 50);
3238 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3239
3240 return resv_clusters;
3241}
3242
3243
3244static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3245{
3246 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3247 sbi->s_cluster_bits;
3248
3249 if (count >= clusters)
3250 return -EINVAL;
3251
3252 atomic64_set(&sbi->s_resv_clusters, count);
3253 return 0;
3254}
3255
2b2d6d01 3256static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3257{
d4c402d9 3258 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3259 struct buffer_head *bh;
617ba13b
MC		 3260 struct ext4_super_block *es = NULL;
3261 struct ext4_sb_info *sbi;
3262 ext4_fsblk_t block;
3263 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3264 ext4_fsblk_t logical_sb_block;
ac27a0ec 3265 unsigned long offset = 0;
ac27a0ec
DK		 3266 unsigned long journal_devnum = 0;
3267 unsigned long def_mount_opts;
3268 struct inode *root;
9f6200bb 3269 char *cp;
0390131b 3270 const char *descr;
dcc7dae3 3271 int ret = -ENOMEM;
281b5995 3272 int blocksize, clustersize;
4ec11028
TT		 3273 unsigned int db_count;
3274 unsigned int i;
281b5995 3275 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3276 __u64 blocks_count;
07aa2ea1 3277 int err = 0;
b3881f74 3278 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3279 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3280
3281 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3282 if (!sbi)
dcc7dae3 3283 goto out_free_orig;
705895b6
PE		 3284
3285 sbi->s_blockgroup_lock =
3286 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3287 if (!sbi->s_blockgroup_lock) {
3288 kfree(sbi);
dcc7dae3 3289 goto out_free_orig;
705895b6 3290 }
ac27a0ec 3291 sb->s_fs_info = sbi;
2c0544b2 3292 sbi->s_sb = sb;
240799cd 3293 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3294 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3295 if (sb->s_bdev->bd_part)
3296 sbi->s_sectors_written_start =
3297 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3298
9f6200bb
TT		 3299 /* Cleanup superblock name */
3300 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3301 *cp = '!';
3302
07aa2ea1 3303 /* -EINVAL is default */
dcc7dae3 3304 ret = -EINVAL;
617ba13b 3305 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3306 if (!blocksize) {
b31e1552 3307 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3308 goto out_fail;
3309 }
3310
3311 /*
617ba13b 3312 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3313 * block sizes. We need to calculate the offset from buffer start.
3314 */
617ba13b 3315 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3316 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3317 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3318 } else {
70bbb3e0 3319 logical_sb_block = sb_block;
ac27a0ec
DK		 3320 }
3321
70bbb3e0 3322 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3323 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3324 goto out_fail;
3325 }
3326 /*
3327 * Note: s_es must be initialized as soon as possible because
617ba13b 3328 * some ext4 macro-instructions depend on its value
ac27a0ec 3329 */
2716b802 3330 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3331 sbi->s_es = es;
3332 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3333 if (sb->s_magic != EXT4_SUPER_MAGIC)
3334 goto cantfind_ext4;
afc32f7e 3335 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3336
feb0ab32
DW		 3337 /* Warn if metadata_csum and gdt_csum are both set. */
3338 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3339 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3340 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3341 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3342 "redundant flags; please run fsck.");
3343
d25425f8
DW		 3344 /* Check for a known checksum algorithm */
3345 if (!ext4_verify_csum_type(sb, es)) {
3346 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3347 "unknown checksum algorithm.");
3348 silent = 1;
3349 goto cantfind_ext4;
3350 }
3351
0441984a
DW		 3352 /* Load the checksum driver */
3353 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3354 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3355 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3356 if (IS_ERR(sbi->s_chksum_driver)) {
3357 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3358 ret = PTR_ERR(sbi->s_chksum_driver);
3359 sbi->s_chksum_driver = NULL;
3360 goto failed_mount;
3361 }
3362 }
3363
a9c47317
DW		 3364 /* Check superblock checksum */
3365 if (!ext4_superblock_csum_verify(sb, es)) {
3366 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3367 "invalid superblock checksum. Run e2fsck?");
3368 silent = 1;
3369 goto cantfind_ext4;
3370 }
3371
3372 /* Precompute checksum seed for all metadata */
3373 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3374 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3375 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3376 sizeof(es->s_uuid));
3377
ac27a0ec
DK		 3378 /* Set defaults before we parse the mount options */
3379 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3380 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3381 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3382 set_opt(sb, DEBUG);
87f26807 3383 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3384 set_opt(sb, GRPID);
617ba13b 3385 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3386 set_opt(sb, NO_UID32);
ea663336 3387 /* xattr user namespace & acls are now defaulted on */
ea663336 3388 set_opt(sb, XATTR_USER);
03010a33 3389#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3390 set_opt(sb, POSIX_ACL);
2e7842b8 3391#endif
617ba13b 3392 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3393 set_opt(sb, JOURNAL_DATA);
617ba13b 3394 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3395 set_opt(sb, ORDERED_DATA);
617ba13b 3396 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3397 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3398
3399 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3400 set_opt(sb, ERRORS_PANIC);
bb4f397a 3401 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3402 set_opt(sb, ERRORS_CONT);
bb4f397a 3403 else
fd8c37ec 3404 set_opt(sb, ERRORS_RO);
8b67f04a 3405 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3406 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3407 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3408 set_opt(sb, DISCARD);
ac27a0ec 3409
08cefc7a
EB		 3410 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3411 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3412 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3413 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3414 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3415
8b67f04a 3416 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3417 set_opt(sb, BARRIER);
ac27a0ec 3418
dd919b98
AK		 3419 /*
3420 * enable delayed allocation by default
3421 * Use -o nodelalloc to turn it off
3422 */
bc0b75f7 3423 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3424 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3425 set_opt(sb, DELALLOC);
dd919b98 3426
51ce6511
LC		 3427 /*
3428 * set default s_li_wait_mult for lazyinit, for the case there is
3429 * no mount option specified.
3430 */
3431 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3432
8b67f04a 3433 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3434 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3435 ext4_msg(sb, KERN_WARNING,
3436 "failed to parse options in superblock: %s",
3437 sbi->s_es->s_mount_opts);
3438 }
5a916be1 3439 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3440 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3441 &journal_ioprio, 0))
ac27a0ec
DK		 3442 goto failed_mount;
3443
56889787
TT		 3444 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3445 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3446 "with data=journal disables delayed "
3447 "allocation and O_DIRECT support!\n");
3448 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3449 ext4_msg(sb, KERN_ERR, "can't mount with "
3450 "both data=journal and delalloc");
3451 goto failed_mount;
3452 }
3453 if (test_opt(sb, DIOREAD_NOLOCK)) {
3454 ext4_msg(sb, KERN_ERR, "can't mount with "
3455 "both data=journal and delalloc");
3456 goto failed_mount;
3457 }
3458 if (test_opt(sb, DELALLOC))
3459 clear_opt(sb, DELALLOC);
3460 }
3461
ac27a0ec 3462 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3463 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3464
617ba13b
MC		 3465 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3466 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3467 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3468 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3469 ext4_msg(sb, KERN_WARNING,
3470 "feature flags set on rev 0 fs, "
3471 "running e2fsck is recommended");
469108ff 3472
2035e776
TT		 3473 if (IS_EXT2_SB(sb)) {
3474 if (ext2_feature_set_ok(sb))
3475 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3476 "using the ext4 subsystem");
3477 else {
3478 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3479 "to feature incompatibilities");
3480 goto failed_mount;
3481 }
3482 }
3483
3484 if (IS_EXT3_SB(sb)) {
3485 if (ext3_feature_set_ok(sb))
3486 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3487 "using the ext4 subsystem");
3488 else {
3489 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3490 "to feature incompatibilities");
3491 goto failed_mount;
3492 }
3493 }
3494
ac27a0ec
DK		 3495 /*
3496 * Check feature flags regardless of the revision level, since we
3497 * previously didn't change the revision level when setting the flags,
3498 * so there is a chance incompat flags are set on a rev 0 filesystem.
3499 */
a13fb1a4 3500 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3501 goto failed_mount;
a13fb1a4 3502
261cb20c 3503 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3504 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3505 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3506 ext4_msg(sb, KERN_ERR,
3507 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3508 goto failed_mount;
3509 }
3510
ac27a0ec 3511 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3512 /* Validate the filesystem blocksize */
3513 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3514 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3515 blocksize);
ac27a0ec
DK		 3516 goto failed_mount;
3517 }
3518
2b2d6d01 3519 brelse(bh);
70bbb3e0
AM		 3520 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3521 offset = do_div(logical_sb_block, blocksize);
3522 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3523 if (!bh) {
b31e1552
ES		 3524 ext4_msg(sb, KERN_ERR,
3525 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3526 goto failed_mount;
3527 }
2716b802 3528 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3529 sbi->s_es = es;
617ba13b 3530 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3531 ext4_msg(sb, KERN_ERR,
3532 "Magic mismatch, very weird!");
ac27a0ec
DK		 3533 goto failed_mount;
3534 }
3535 }
3536
a13fb1a4
ES		 3537 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3538 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3539 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3540 has_huge_files);
3541 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3542
617ba13b
MC		 3543 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3544 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3545 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3546 } else {
3547 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3548 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3549 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3550 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3551 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3552 ext4_msg(sb, KERN_ERR,
3553 "unsupported inode size: %d",
2b2d6d01 3554 sbi->s_inode_size);
ac27a0ec
DK		 3555 goto failed_mount;
3556 }
ef7f3835
KS		 3557 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3558 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3559 }
0b8e58a1 3560
0d1ee42f
AR		 3561 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3562 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3563 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3564 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3565 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3566 ext4_msg(sb, KERN_ERR,
3567 "unsupported descriptor size %lu",
0d1ee42f
AR		 3568 sbi->s_desc_size);
3569 goto failed_mount;
3570 }
3571 } else
3572 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3573
ac27a0ec 3574 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3575 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3576 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3577 goto cantfind_ext4;
0b8e58a1 3578
617ba13b 3579 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3580 if (sbi->s_inodes_per_block == 0)
617ba13b 3581 goto cantfind_ext4;
ac27a0ec
DK		 3582 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3583 sbi->s_inodes_per_block;
0d1ee42f 3584 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3585 sbi->s_sbh = bh;
3586 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3587 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3588 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3589
68911009
LC		 3590 /* Do we have standard group size of blocksize * 8 blocks ? */
3591 if (sbi->s_blocks_per_group == blocksize << 3)
3592 set_opt2(sb, STD_GROUP_SIZE);
3593
2b2d6d01 3594 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3595 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3596 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3597 i = le32_to_cpu(es->s_flags);
3598 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3599 sbi->s_hash_unsigned = 3;
3600 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3601#ifdef __CHAR_UNSIGNED__
3602 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3603 sbi->s_hash_unsigned = 3;
3604#else
3605 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3606#endif
f99b2589 3607 }
ac27a0ec 3608
281b5995
TT		 3609 /* Handle clustersize */
3610 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3611 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3612 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3613 if (has_bigalloc) {
3614 if (clustersize < blocksize) {
3615 ext4_msg(sb, KERN_ERR,
3616 "cluster size (%d) smaller than "
3617 "block size (%d)", clustersize, blocksize);
3618 goto failed_mount;
3619 }
3620 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3621 le32_to_cpu(es->s_log_block_size);
3622 sbi->s_clusters_per_group =
3623 le32_to_cpu(es->s_clusters_per_group);
3624 if (sbi->s_clusters_per_group > blocksize * 8) {
3625 ext4_msg(sb, KERN_ERR,
3626 "#clusters per group too big: %lu",
3627 sbi->s_clusters_per_group);
3628 goto failed_mount;
3629 }
3630 if (sbi->s_blocks_per_group !=
3631 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3632 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3633 "clusters per group (%lu) inconsistent",
3634 sbi->s_blocks_per_group,
3635 sbi->s_clusters_per_group);
3636 goto failed_mount;
3637 }
3638 } else {
3639 if (clustersize != blocksize) {
3640 ext4_warning(sb, "fragment/cluster size (%d) != "
3641 "block size (%d)", clustersize,
3642 blocksize);
3643 clustersize = blocksize;
3644 }
3645 if (sbi->s_blocks_per_group > blocksize * 8) {
3646 ext4_msg(sb, KERN_ERR,
3647 "#blocks per group too big: %lu",
3648 sbi->s_blocks_per_group);
3649 goto failed_mount;
3650 }
3651 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3652 sbi->s_cluster_bits = 0;
ac27a0ec 3653 }
281b5995
TT		 3654 sbi->s_cluster_ratio = clustersize / blocksize;
3655
ac27a0ec 3656 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3657 ext4_msg(sb, KERN_ERR,
3658 "#inodes per group too big: %lu",
2b2d6d01 3659 sbi->s_inodes_per_group);
ac27a0ec
DK		 3660 goto failed_mount;
3661 }
3662
bf43d84b
ES		 3663 /*
3664 * Test whether we have more sectors than will fit in sector_t,
3665 * and whether the max offset is addressable by the page cache.
3666 */
5a9ae68a 3667 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3668 ext4_blocks_count(es));
5a9ae68a 3669 if (err) {
b31e1552 3670 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3671 " too large to mount safely on this system");
ac27a0ec 3672 if (sizeof(sector_t) < 8)
90c699a9 3673 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3674 goto failed_mount;
3675 }
3676
617ba13b
MC		 3677 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3678 goto cantfind_ext4;
e7c95593 3679
0f2ddca6
FTN		 3680 /* check blocks count against device size */
3681 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3682 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3683 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3684 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3685 ext4_blocks_count(es), blocks_count);
3686 goto failed_mount;
3687 }
3688
0b8e58a1
AD		 3689 /*
3690 * It makes no sense for the first data block to be beyond the end
3691 * of the filesystem.
3692 */
3693 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3694 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3695 "block %u is beyond end of filesystem (%llu)",
3696 le32_to_cpu(es->s_first_data_block),
3697 ext4_blocks_count(es));
e7c95593
ES		 3698 goto failed_mount;
3699 }
bd81d8ee
LV		 3700 blocks_count = (ext4_blocks_count(es) -
3701 le32_to_cpu(es->s_first_data_block) +
3702 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3703 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3704 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3705 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3706 "(block count %llu, first data block %u, "
b31e1552 3707 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3708 ext4_blocks_count(es),
3709 le32_to_cpu(es->s_first_data_block),
3710 EXT4_BLOCKS_PER_GROUP(sb));
3711 goto failed_mount;
3712 }
bd81d8ee 3713 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3714 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3715 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3716 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3717 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3718 sbi->s_group_desc = ext4_kvmalloc(db_count *
3719 sizeof(struct buffer_head *),
3720 GFP_KERNEL);
ac27a0ec 3721 if (sbi->s_group_desc == NULL) {
b31e1552 3722 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3723 ret = -ENOMEM;
ac27a0ec
DK		 3724 goto failed_mount;
3725 }
3726
9f6200bb
TT		 3727 if (ext4_proc_root)
3728 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3729
66acdcf4
TT		 3730 if (sbi->s_proc)
3731 proc_create_data("options", S_IRUGO, sbi->s_proc,
3732 &ext4_seq_options_fops, sb);
3733
705895b6 3734 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3735
3736 for (i = 0; i < db_count; i++) {
70bbb3e0 3737 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3738 sbi->s_group_desc[i] = sb_bread(sb, block);
3739 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3740 ext4_msg(sb, KERN_ERR,
3741 "can't read group descriptor %d", i);
ac27a0ec
DK		 3742 db_count = i;
3743 goto failed_mount2;
3744 }
3745 }
bfff6873 3746 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3747 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3748 goto failed_mount2;
3749 }
772cb7c8
JS		 3750 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3751 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3752 ext4_msg(sb, KERN_ERR,
3753 "unable to initialize "
3754 "flex_bg meta info!");
772cb7c8
JS		 3755 goto failed_mount2;
3756 }
3757
ac27a0ec
DK		 3758 sbi->s_gdb_count = db_count;
3759 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3760 spin_lock_init(&sbi->s_next_gen_lock);
3761
04496411
TM		 3762 init_timer(&sbi->s_err_report);
3763 sbi->s_err_report.function = print_daily_error_info;
3764 sbi->s_err_report.data = (unsigned long) sb;
3765
a75ae78f
DM		 3766 /* Register extent status tree shrinker */
3767 ext4_es_register_shrinker(sb);
3768
57042651 3769 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3770 ext4_count_free_clusters(sb));
ce7e010a
TT		 3771 if (!err) {
3772 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3773 ext4_count_free_inodes(sb));
3774 }
3775 if (!err) {
3776 err = percpu_counter_init(&sbi->s_dirs_counter,
3777 ext4_count_dirs(sb));
3778 }
3779 if (!err) {
57042651 3780 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3781 }
1ac6466f
TT		 3782 if (!err) {
3783 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3784 }
ce7e010a
TT		 3785 if (err) {
3786 ext4_msg(sb, KERN_ERR, "insufficient memory");
3787 goto failed_mount3;
3788 }
3789
c9de560d 3790 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3791 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3792 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3793
ac27a0ec
DK		 3794 /*
3795 * set up enough so that it can read an inode
3796 */
9ca92389
TT		 3797 if (!test_opt(sb, NOLOAD) &&
3798 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3799 sb->s_op = &ext4_sops;
3800 else
3801 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3802 sb->s_export_op = &ext4_export_ops;
3803 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3804#ifdef CONFIG_QUOTA
617ba13b 3805 sb->dq_op = &ext4_quota_operations;
262b4662 3806 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3807 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3808 else
3809 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3810#endif
f2fa2ffc
AK		 3811 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3812
ac27a0ec 3813 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3814 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3815
3816 sb->s_root = NULL;
3817
3818 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3819 EXT4_HAS_INCOMPAT_FEATURE(sb,
3820 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3821
c5e06d10
JL		 3822 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3823 !(sb->s_flags & MS_RDONLY))
3824 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3825 goto failed_mount3;
3826
ac27a0ec
DK		 3827 /*
3828 * The first inode we look at is the journal inode. Don't try
3829 * root first: it may be modified in the journal!
3830 */
3831 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3832 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3833 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3834 goto failed_mount3;
0390131b
FM		 3835 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3836 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3837 ext4_msg(sb, KERN_ERR, "required journal recovery "
3838 "suppressed and not mounted read-only");
744692dc 3839 goto failed_mount_wq;
ac27a0ec 3840 } else {
fd8c37ec 3841 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3842 sbi->s_journal = NULL;
3843 needs_recovery = 0;
3844 goto no_journal;
ac27a0ec
DK		 3845 }
3846
f32aaf2d 3847 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3848 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3849 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3850 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3851 goto failed_mount_wq;
eb40a09c
JS		 3852 }
3853
25ed6e8a
DW		 3854 if (!set_journal_csum_feature_set(sb)) {
3855 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3856 "feature set");
3857 goto failed_mount_wq;
d4da6c9c 3858 }
818d276c 3859
ac27a0ec
DK		 3860 /* We have now updated the journal if required, so we can
3861 * validate the data journaling mode. */
3862 switch (test_opt(sb, DATA_FLAGS)) {
3863 case 0:
3864 /* No mode set, assume a default based on the journal
63f57933
AM		 3865 * capabilities: ORDERED_DATA if the journal can
3866 * cope, else JOURNAL_DATA
3867 */
dab291af
MC		 3868 if (jbd2_journal_check_available_features
3869 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3870 set_opt(sb, ORDERED_DATA);
ac27a0ec 3871 else
fd8c37ec 3872 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3873 break;
3874
617ba13b
MC		 3875 case EXT4_MOUNT_ORDERED_DATA:
3876 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3877 if (!jbd2_journal_check_available_features
3878 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3879 ext4_msg(sb, KERN_ERR, "Journal does not support "
3880 "requested data journaling mode");
744692dc 3881 goto failed_mount_wq;
ac27a0ec
DK		 3882 }
3883 default:
3884 break;
3885 }
b3881f74 3886 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3887
18aadd47
BJ		 3888 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3889
ce7e010a
TT		 3890 /*
3891 * The journal may have updated the bg summary counts, so we
3892 * need to update the global counters.
3893 */
57042651 3894 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3895 ext4_count_free_clusters(sb));
ce7e010a
TT		 3896 percpu_counter_set(&sbi->s_freeinodes_counter,
3897 ext4_count_free_inodes(sb));
3898 percpu_counter_set(&sbi->s_dirs_counter,
3899 ext4_count_dirs(sb));
57042651 3900 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3901
ce7e010a 3902no_journal:
952fc18e
TT		 3903 /*
3904 * Get the # of file system overhead blocks from the
3905 * superblock if present.
3906 */
3907 if (es->s_overhead_clusters)
3908 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3909 else {
07aa2ea1
LC		 3910 err = ext4_calculate_overhead(sb);
3911 if (err)
952fc18e
TT		 3912 goto failed_mount_wq;
3913 }
3914
fd89d5f2
TH		 3915 /*
3916 * The maximum number of concurrent works can be high and
3917 * concurrency isn't really necessary. Limit it to 1.
3918 */
3919 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3920 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3921 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3922 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3923 ret = -ENOMEM;
4c0425ff
MC		 3924 goto failed_mount_wq;
3925 }
3926
ac27a0ec 3927 /*
dab291af 3928 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3929 * so we can safely mount the rest of the filesystem now.
3930 */
3931
1d1fe1ee
DH		 3932 root = ext4_iget(sb, EXT4_ROOT_INO);
3933 if (IS_ERR(root)) {
b31e1552 3934 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3935 ret = PTR_ERR(root);
32a9bb57 3936 root = NULL;
ac27a0ec
DK		 3937 goto failed_mount4;
3938 }
3939 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3940 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3941 iput(root);
ac27a0ec
DK		 3942 goto failed_mount4;
3943 }
48fde701 3944 sb->s_root = d_make_root(root);
1d1fe1ee 3945 if (!sb->s_root) {
b31e1552 3946 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3947 ret = -ENOMEM;
3948 goto failed_mount4;
3949 }
ac27a0ec 3950
7e84b621
ES		 3951 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3952 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3953
3954 /* determine the minimum size of new large inodes, if present */
3955 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3956 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3957 EXT4_GOOD_OLD_INODE_SIZE;
3958 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3959 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3960 if (sbi->s_want_extra_isize <
3961 le16_to_cpu(es->s_want_extra_isize))
3962 sbi->s_want_extra_isize =
3963 le16_to_cpu(es->s_want_extra_isize);
3964 if (sbi->s_want_extra_isize <
3965 le16_to_cpu(es->s_min_extra_isize))
3966 sbi->s_want_extra_isize =
3967 le16_to_cpu(es->s_min_extra_isize);
3968 }
3969 }
3970 /* Check if enough inode space is available */
3971 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3972 sbi->s_inode_size) {
3973 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3974 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3975 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3976 "available");
ef7f3835
KS		 3977 }
3978
27dd4385
LC		 3979 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
3980 if (err) {
3981 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
3982 "reserved pool", ext4_calculate_resv_clusters(sbi));
3983 goto failed_mount4a;
3984 }
3985
6fd058f7
TT		 3986 err = ext4_setup_system_zone(sb);
3987 if (err) {
b31e1552 3988 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3989 "zone (%d)", err);
94bf608a 3990 goto failed_mount4a;
6fd058f7
TT		 3991 }
3992
c2774d84 3993 ext4_ext_init(sb);
9d99012f 3994 err = ext4_mb_init(sb);
c2774d84 3995 if (err) {
421f91d2 3996 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3997 err);
dcf2d804 3998 goto failed_mount5;
c2774d84
AK		 3999 }
4000
bfff6873
LC		 4001 err = ext4_register_li_request(sb, first_not_zeroed);
4002 if (err)
dcf2d804 4003 goto failed_mount6;
bfff6873 4004
3197ebdb
TT		 4005 sbi->s_kobj.kset = ext4_kset;
4006 init_completion(&sbi->s_kobj_unregister);
4007 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4008 "%s", sb->s_id);
dcf2d804
TM		 4009 if (err)
4010 goto failed_mount7;
3197ebdb 4011
9b2ff357
JK		 4012#ifdef CONFIG_QUOTA
4013 /* Enable quota usage during mount. */
4014 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4015 !(sb->s_flags & MS_RDONLY)) {
4016 err = ext4_enable_quotas(sb);
4017 if (err)
4018 goto failed_mount8;
4019 }
4020#endif /* CONFIG_QUOTA */
4021
617ba13b
MC		 4022 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4023 ext4_orphan_cleanup(sb, es);
4024 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4025 if (needs_recovery) {
b31e1552 4026 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4027 ext4_mark_recovery_complete(sb, es);
4028 }
4029 if (EXT4_SB(sb)->s_journal) {
4030 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4031 descr = " journalled data mode";
4032 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4033 descr = " ordered data mode";
4034 else
4035 descr = " writeback data mode";
4036 } else
4037 descr = "out journal";
4038
79add3a3
LC		 4039 if (test_opt(sb, DISCARD)) {
4040 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4041 if (!blk_queue_discard(q))
4042 ext4_msg(sb, KERN_WARNING,
4043 "mounting with \"discard\" option, but "
4044 "the device does not support discard");
4045 }
4046
d4c402d9 4047 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4048 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4049 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4050
66e61a9e
TT		 4051 if (es->s_error_count)
4052 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4053
d4c402d9 4054 kfree(orig_data);
ac27a0ec
DK		 4055 return 0;
4056
617ba13b 4057cantfind_ext4:
ac27a0ec 4058 if (!silent)
b31e1552 4059 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4060 goto failed_mount;
4061
72ba7450
TT		 4062#ifdef CONFIG_QUOTA
4063failed_mount8:
4064 kobject_del(&sbi->s_kobj);
4065#endif
dcf2d804
TM		 4066failed_mount7:
4067 ext4_unregister_li_request(sb);
4068failed_mount6:
dcf2d804 4069 ext4_mb_release(sb);
94bf608a
AV		 4070failed_mount5:
4071 ext4_ext_release(sb);
dcf2d804 4072 ext4_release_system_zone(sb);
94bf608a
AV		 4073failed_mount4a:
4074 dput(sb->s_root);
32a9bb57 4075 sb->s_root = NULL;
94bf608a 4076failed_mount4:
b31e1552 4077 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4078 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4079failed_mount_wq:
0390131b
FM		 4080 if (sbi->s_journal) {
4081 jbd2_journal_destroy(sbi->s_journal);
4082 sbi->s_journal = NULL;
4083 }
ac27a0ec 4084failed_mount3:
a75ae78f 4085 ext4_es_unregister_shrinker(sb);
04496411 4086 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4087 if (sbi->s_flex_groups)
4088 ext4_kvfree(sbi->s_flex_groups);
57042651 4089 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4090 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4091 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4092 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4093 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4094 if (sbi->s_mmp_tsk)
4095 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4096failed_mount2:
4097 for (i = 0; i < db_count; i++)
4098 brelse(sbi->s_group_desc[i]);
f18a5f21 4099 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4100failed_mount:
0441984a
DW		 4101 if (sbi->s_chksum_driver)
4102 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4103 if (sbi->s_proc) {
66acdcf4 4104 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4105 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4106 }
ac27a0ec
DK		 4107#ifdef CONFIG_QUOTA
4108 for (i = 0; i < MAXQUOTAS; i++)
4109 kfree(sbi->s_qf_names[i]);
4110#endif
617ba13b 4111 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4112 brelse(bh);
4113out_fail:
4114 sb->s_fs_info = NULL;
f6830165 4115 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4116 kfree(sbi);
dcc7dae3 4117out_free_orig:
d4c402d9 4118 kfree(orig_data);
07aa2ea1 4119 return err ? err : ret;
ac27a0ec
DK		 4120}
4121
4122/*
4123 * Setup any per-fs journal parameters now. We'll do this both on
4124 * initial mount, once the journal has been initialised but before we've
4125 * done any recovery; and again on any subsequent remount.
4126 */
617ba13b 4127static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4128{
617ba13b 4129 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4130
30773840
TT		 4131 journal->j_commit_interval = sbi->s_commit_interval;
4132 journal->j_min_batch_time = sbi->s_min_batch_time;
4133 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4134
a931da6a 4135 write_lock(&journal->j_state_lock);
ac27a0ec 4136 if (test_opt(sb, BARRIER))
dab291af 4137 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4138 else
dab291af 4139 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4140 if (test_opt(sb, DATA_ERR_ABORT))
4141 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4142 else
4143 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4144 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4145}
4146
617ba13b 4147static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4148 unsigned int journal_inum)
4149{
4150 struct inode *journal_inode;
4151 journal_t *journal;
4152
0390131b
FM		 4153 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4154
ac27a0ec
DK		 4155 /* First, test for the existence of a valid inode on disk. Bad
4156 * things happen if we iget() an unused inode, as the subsequent
4157 * iput() will try to delete it. */
4158
1d1fe1ee
DH		 4159 journal_inode = ext4_iget(sb, journal_inum);
4160 if (IS_ERR(journal_inode)) {
b31e1552 4161 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4162 return NULL;
4163 }
4164 if (!journal_inode->i_nlink) {
4165 make_bad_inode(journal_inode);
4166 iput(journal_inode);
b31e1552 4167 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4168 return NULL;
4169 }
4170
e5f8eab8 4171 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4172 journal_inode, journal_inode->i_size);
1d1fe1ee 4173 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4174 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4175 iput(journal_inode);
4176 return NULL;
4177 }
4178
dab291af 4179 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4180 if (!journal) {
b31e1552 4181 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4182 iput(journal_inode);
4183 return NULL;
4184 }
4185 journal->j_private = sb;
617ba13b 4186 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4187 return journal;
4188}
4189
617ba13b 4190static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4191 dev_t j_dev)
4192{
2b2d6d01 4193 struct buffer_head *bh;
ac27a0ec 4194 journal_t *journal;
617ba13b
MC		 4195 ext4_fsblk_t start;
4196 ext4_fsblk_t len;
ac27a0ec 4197 int hblock, blocksize;
617ba13b 4198 ext4_fsblk_t sb_block;
ac27a0ec 4199 unsigned long offset;
2b2d6d01 4200 struct ext4_super_block *es;
ac27a0ec
DK		 4201 struct block_device *bdev;
4202
0390131b
FM		 4203 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4204
b31e1552 4205 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4206 if (bdev == NULL)
4207 return NULL;
4208
ac27a0ec 4209 blocksize = sb->s_blocksize;
e1defc4f 4210 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4211 if (blocksize < hblock) {
b31e1552
ES		 4212 ext4_msg(sb, KERN_ERR,
4213 "blocksize too small for journal device");
ac27a0ec
DK		 4214 goto out_bdev;
4215 }
4216
617ba13b
MC		 4217 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4218 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4219 set_blocksize(bdev, blocksize);
4220 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4221 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4222 "external journal");
ac27a0ec
DK		 4223 goto out_bdev;
4224 }
4225
2716b802 4226 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4227 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4228 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4229 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4230 ext4_msg(sb, KERN_ERR, "external journal has "
4231 "bad superblock");
ac27a0ec
DK		 4232 brelse(bh);
4233 goto out_bdev;
4234 }
4235
617ba13b 4236 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4237 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4238 brelse(bh);
4239 goto out_bdev;
4240 }
4241
bd81d8ee 4242 len = ext4_blocks_count(es);
ac27a0ec
DK		 4243 start = sb_block + 1;
4244 brelse(bh); /* we're done with the superblock */
4245
dab291af 4246 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4247 start, len, blocksize);
4248 if (!journal) {
b31e1552 4249 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4250 goto out_bdev;
4251 }
4252 journal->j_private = sb;
9f203507 4253 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4254 wait_on_buffer(journal->j_sb_buffer);
4255 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4256 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4257 goto out_journal;
4258 }
4259 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4260 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4261 "user (unsupported) - %d",
ac27a0ec
DK		 4262 be32_to_cpu(journal->j_superblock->s_nr_users));
4263 goto out_journal;
4264 }
617ba13b
MC		 4265 EXT4_SB(sb)->journal_bdev = bdev;
4266 ext4_init_journal_params(sb, journal);
ac27a0ec 4267 return journal;
0b8e58a1 4268
ac27a0ec 4269out_journal:
dab291af 4270 jbd2_journal_destroy(journal);
ac27a0ec 4271out_bdev:
617ba13b 4272 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4273 return NULL;
4274}
4275
617ba13b
MC		 4276static int ext4_load_journal(struct super_block *sb,
4277 struct ext4_super_block *es,
ac27a0ec
DK		 4278 unsigned long journal_devnum)
4279{
4280 journal_t *journal;
4281 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4282 dev_t journal_dev;
4283 int err = 0;
4284 int really_read_only;
4285
0390131b
FM		 4286 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4287
ac27a0ec
DK		 4288 if (journal_devnum &&
4289 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4290 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4291 "numbers have changed");
ac27a0ec
DK		 4292 journal_dev = new_decode_dev(journal_devnum);
4293 } else
4294 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4295
4296 really_read_only = bdev_read_only(sb->s_bdev);
4297
4298 /*
4299 * Are we loading a blank journal or performing recovery after a
4300 * crash? For recovery, we need to check in advance whether we
4301 * can get read-write access to the device.
4302 */
617ba13b 4303 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4304 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4305 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4306 "required on readonly filesystem");
ac27a0ec 4307 if (really_read_only) {
b31e1552
ES		 4308 ext4_msg(sb, KERN_ERR, "write access "
4309 "unavailable, cannot proceed");
ac27a0ec
DK		 4310 return -EROFS;
4311 }
b31e1552
ES		 4312 ext4_msg(sb, KERN_INFO, "write access will "
4313 "be enabled during recovery");
ac27a0ec
DK		 4314 }
4315 }
4316
4317 if (journal_inum && journal_dev) {
b31e1552
ES		 4318 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4319 "and inode journals!");
ac27a0ec
DK		 4320 return -EINVAL;
4321 }
4322
4323 if (journal_inum) {
617ba13b 4324 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4325 return -EINVAL;
4326 } else {
617ba13b 4327 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4328 return -EINVAL;
4329 }
4330
90576c0b 4331 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4332 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4333
617ba13b 4334 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4335 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4336 if (!err) {
4337 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4338 if (save)
4339 memcpy(save, ((char *) es) +
4340 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4341 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4342 if (save)
4343 memcpy(((char *) es) + EXT4_S_ERR_START,
4344 save, EXT4_S_ERR_LEN);
4345 kfree(save);
4346 }
ac27a0ec
DK		 4347
4348 if (err) {
b31e1552 4349 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4350 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4351 return err;
4352 }
4353
617ba13b
MC		 4354 EXT4_SB(sb)->s_journal = journal;
4355 ext4_clear_journal_err(sb, es);
ac27a0ec 4356
c41303ce 4357 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4358 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4359 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4360
4361 /* Make sure we flush the recovery flag to disk. */
e2d67052 4362 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4363 }
4364
4365 return 0;
4366}
4367
e2d67052 4368static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4369{
e2d67052 4370 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4371 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4372 int error = 0;
ac27a0ec 4373
7c2e7087 4374 if (!sbh || block_device_ejected(sb))
c4be0c1d 4375 return error;
914258bf
TT		 4376 if (buffer_write_io_error(sbh)) {
4377 /*
4378 * Oh, dear. A previous attempt to write the
4379 * superblock failed. This could happen because the
4380 * USB device was yanked out. Or it could happen to
4381 * be a transient write error and maybe the block will
4382 * be remapped. Nothing we can do but to retry the
4383 * write and hope for the best.
4384 */
b31e1552
ES		 4385 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4386 "superblock detected");
914258bf
TT		 4387 clear_buffer_write_io_error(sbh);
4388 set_buffer_uptodate(sbh);
4389 }
71290b36
TT		 4390 /*
4391 * If the file system is mounted read-only, don't update the
4392 * superblock write time. This avoids updating the superblock
4393 * write time when we are mounting the root file system
4394 * read/only but we need to replay the journal; at that point,
4395 * for people who are east of GMT and who make their clock
4396 * tick in localtime for Windows bug-for-bug compatibility,
4397 * the clock is set in the future, and this will cause e2fsck
4398 * to complain and force a full file system check.
4399 */
4400 if (!(sb->s_flags & MS_RDONLY))
4401 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4402 if (sb->s_bdev->bd_part)
4403 es->s_kbytes_written =
4404 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4405 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4406 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4407 else
4408 es->s_kbytes_written =
4409 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4410 ext4_free_blocks_count_set(es,
4411 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4412 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4413 es->s_free_inodes_count =
4414 cpu_to_le32(percpu_counter_sum_positive(
4415 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4416 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4417 ext4_superblock_csum_set(sb);
ac27a0ec 4418 mark_buffer_dirty(sbh);
914258bf 4419 if (sync) {
c4be0c1d
TS		 4420 error = sync_dirty_buffer(sbh);
4421 if (error)
4422 return error;
4423
4424 error = buffer_write_io_error(sbh);
4425 if (error) {
b31e1552
ES		 4426 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4427 "superblock");
914258bf
TT		 4428 clear_buffer_write_io_error(sbh);
4429 set_buffer_uptodate(sbh);
4430 }
4431 }
c4be0c1d 4432 return error;
ac27a0ec
DK		 4433}
4434
ac27a0ec
DK		 4435/*
4436 * Have we just finished recovery? If so, and if we are mounting (or
4437 * remounting) the filesystem readonly, then we will end up with a
4438 * consistent fs on disk. Record that fact.
4439 */
2b2d6d01
TT		 4440static void ext4_mark_recovery_complete(struct super_block *sb,
4441 struct ext4_super_block *es)
ac27a0ec 4442{
617ba13b 4443 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4444
0390131b
FM		 4445 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4446 BUG_ON(journal != NULL);
4447 return;
4448 }
dab291af 4449 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4450 if (jbd2_journal_flush(journal) < 0)
4451 goto out;
4452
617ba13b 4453 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4454 sb->s_flags & MS_RDONLY) {
617ba13b 4455 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4456 ext4_commit_super(sb, 1);
ac27a0ec 4457 }
7ffe1ea8
HK		 4458
4459out:
dab291af 4460 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4461}
4462
4463/*
4464 * If we are mounting (or read-write remounting) a filesystem whose journal
4465 * has recorded an error from a previous lifetime, move that error to the
4466 * main filesystem now.
4467 */
2b2d6d01
TT		 4468static void ext4_clear_journal_err(struct super_block *sb,
4469 struct ext4_super_block *es)
ac27a0ec
DK		 4470{
4471 journal_t *journal;
4472 int j_errno;
4473 const char *errstr;
4474
0390131b
FM		 4475 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4476
617ba13b 4477 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4478
4479 /*
4480 * Now check for any error status which may have been recorded in the
617ba13b 4481 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4482 */
4483
dab291af 4484 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4485 if (j_errno) {
4486 char nbuf[16];
4487
617ba13b 4488 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4489 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4490 "from previous mount: %s", errstr);
12062ddd 4491 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4492
617ba13b
MC		 4493 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4494 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4495 ext4_commit_super(sb, 1);
ac27a0ec 4496
dab291af 4497 jbd2_journal_clear_err(journal);
d796c52e 4498 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4499 }
4500}
4501
4502/*
4503 * Force the running and committing transactions to commit,
4504 * and wait on the commit.
4505 */
617ba13b 4506int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4507{
4508 journal_t *journal;
ac27a0ec
DK		 4509
4510 if (sb->s_flags & MS_RDONLY)
4511 return 0;
4512
617ba13b 4513 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4514 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4515}
4516
617ba13b 4517static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4518{
14ce0cb4 4519 int ret = 0;
9eddacf9 4520 tid_t target;
8d5d02e6 4521 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4522
9bffad1e 4523 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4524 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4525 /*
4526 * Writeback quota in non-journalled quota case - journalled quota has
4527 * no dirty dquots
4528 */
4529 dquot_writeback_dquots(sb, -1);
8d5d02e6 4530 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4531 if (wait)
8d5d02e6 4532 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4533 }
14ce0cb4 4534 return ret;
ac27a0ec
DK		 4535}
4536
4537/*
4538 * LVM calls this function before a (read-only) snapshot is created. This
4539 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4540 *
4541 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4542 * state independently. It relies on upper layer to stop all data & metadata
4543 * modifications.
ac27a0ec 4544 */
c4be0c1d 4545static int ext4_freeze(struct super_block *sb)
ac27a0ec 4546{
c4be0c1d
TS		 4547 int error = 0;
4548 journal_t *journal;
ac27a0ec 4549
9ca92389
TT		 4550 if (sb->s_flags & MS_RDONLY)
4551 return 0;
ac27a0ec 4552
9ca92389 4553 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4554
9ca92389
TT		 4555 /* Now we set up the journal barrier. */
4556 jbd2_journal_lock_updates(journal);
ac27a0ec 4557
9ca92389
TT		 4558 /*
4559 * Don't clear the needs_recovery flag if we failed to flush
4560 * the journal.
4561 */
4562 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4563 if (error < 0)
4564 goto out;
9ca92389
TT		 4565
4566 /* Journal blocked and flushed, clear needs_recovery flag. */
4567 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4568 error = ext4_commit_super(sb, 1);
6b0310fb 4569out:
8e8ad8a5 4570 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4571 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4572 return error;
ac27a0ec
DK		 4573}
4574
4575/*
4576 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4577 * flag here, even though the filesystem is not technically dirty yet.
4578 */
c4be0c1d 4579static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4580{
9ca92389
TT		 4581 if (sb->s_flags & MS_RDONLY)
4582 return 0;
4583
9ca92389
TT		 4584 /* Reset the needs_recovery flag before the fs is unlocked. */
4585 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4586 ext4_commit_super(sb, 1);
c4be0c1d 4587 return 0;
ac27a0ec
DK		 4588}
4589
673c6100
TT		 4590/*
4591 * Structure to save mount options for ext4_remount's benefit
4592 */
4593struct ext4_mount_options {
4594 unsigned long s_mount_opt;
a2595b8a 4595 unsigned long s_mount_opt2;
08cefc7a
EB		 4596 kuid_t s_resuid;
4597 kgid_t s_resgid;
673c6100
TT		 4598 unsigned long s_commit_interval;
4599 u32 s_min_batch_time, s_max_batch_time;
4600#ifdef CONFIG_QUOTA
4601 int s_jquota_fmt;
4602 char *s_qf_names[MAXQUOTAS];
4603#endif
4604};
4605
2b2d6d01 4606static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4607{
2b2d6d01 4608 struct ext4_super_block *es;
617ba13b 4609 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4610 unsigned long old_sb_flags;
617ba13b 4611 struct ext4_mount_options old_opts;
c79d967d 4612 int enable_quota = 0;
8a266467 4613 ext4_group_t g;
b3881f74 4614 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4615 int err = 0;
ac27a0ec 4616#ifdef CONFIG_QUOTA
03dafb5f 4617 int i, j;
ac27a0ec 4618#endif
d4c402d9 4619 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4620
4621 /* Store the original options */
4622 old_sb_flags = sb->s_flags;
4623 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4624 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4625 old_opts.s_resuid = sbi->s_resuid;
4626 old_opts.s_resgid = sbi->s_resgid;
4627 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4628 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4629 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4630#ifdef CONFIG_QUOTA
4631 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4632 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4633 if (sbi->s_qf_names[i]) {
4634 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4635 GFP_KERNEL);
4636 if (!old_opts.s_qf_names[i]) {
4637 for (j = 0; j < i; j++)
4638 kfree(old_opts.s_qf_names[j]);
3e36a163 4639 kfree(orig_data);
03dafb5f
CG		 4640 return -ENOMEM;
4641 }
4642 } else
4643 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4644#endif
b3881f74
TT		 4645 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4646 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4647
4648 /*
4649 * Allow the "check" option to be passed as a remount option.
4650 */
661aa520 4651 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4652 err = -EINVAL;
4653 goto restore_opts;
4654 }
4655
4ab2f15b 4656 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4657 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4658
4659 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4660 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4661
4662 es = sbi->s_es;
4663
b3881f74 4664 if (sbi->s_journal) {
0390131b 4665 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4666 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4667 }
ac27a0ec 4668
661aa520 4669 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4670 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4671 err = -EROFS;
4672 goto restore_opts;
4673 }
4674
4675 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4676 err = dquot_suspend(sb, -1);
4677 if (err < 0)
c79d967d 4678 goto restore_opts;
c79d967d 4679
ac27a0ec
DK		 4680 /*
4681 * First of all, the unconditional stuff we have to do
4682 * to disable replay of the journal when we next remount
4683 */
4684 sb->s_flags |= MS_RDONLY;
4685
4686 /*
4687 * OK, test if we are remounting a valid rw partition
4688 * readonly, and if so set the rdonly flag and then
4689 * mark the partition as valid again.
4690 */
617ba13b
MC		 4691 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4692 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4693 es->s_state = cpu_to_le16(sbi->s_mount_state);
4694
a63c9eb2 4695 if (sbi->s_journal)
0390131b 4696 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4697 } else {
a13fb1a4
ES		 4698 /* Make sure we can mount this feature set readwrite */
4699 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4700 err = -EROFS;
4701 goto restore_opts;
4702 }
8a266467
TT		 4703 /*
4704 * Make sure the group descriptor checksums
0b8e58a1 4705 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4706 */
4707 for (g = 0; g < sbi->s_groups_count; g++) {
4708 struct ext4_group_desc *gdp =
4709 ext4_get_group_desc(sb, g, NULL);
4710
feb0ab32 4711 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4712 ext4_msg(sb, KERN_ERR,
4713 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4714 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4715 le16_to_cpu(gdp->bg_checksum));
4716 err = -EINVAL;
4717 goto restore_opts;
4718 }
4719 }
4720
ead6596b
ES		 4721 /*
4722 * If we have an unprocessed orphan list hanging
4723 * around from a previously readonly bdev mount,
4724 * require a full umount/remount for now.
4725 */
4726 if (es->s_last_orphan) {
b31e1552 4727 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4728 "remount RDWR because of unprocessed "
4729 "orphan inode list. Please "
b31e1552 4730 "umount/remount instead");
ead6596b
ES		 4731 err = -EINVAL;
4732 goto restore_opts;
4733 }
4734
ac27a0ec
DK		 4735 /*
4736 * Mounting a RDONLY partition read-write, so reread
4737 * and store the current valid flag. (It may have
4738 * been changed by e2fsck since we originally mounted
4739 * the partition.)
4740 */
0390131b
FM		 4741 if (sbi->s_journal)
4742 ext4_clear_journal_err(sb, es);
ac27a0ec 4743 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4744 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4745 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4746 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4747 EXT4_FEATURE_INCOMPAT_MMP))
4748 if (ext4_multi_mount_protect(sb,
4749 le64_to_cpu(es->s_mmp_block))) {
4750 err = -EROFS;
4751 goto restore_opts;
4752 }
c79d967d 4753 enable_quota = 1;
ac27a0ec
DK		 4754 }
4755 }
bfff6873
LC		 4756
4757 /*
4758 * Reinitialize lazy itable initialization thread based on
4759 * current settings
4760 */
4761 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4762 ext4_unregister_li_request(sb);
4763 else {
4764 ext4_group_t first_not_zeroed;
4765 first_not_zeroed = ext4_has_uninit_itable(sb);
4766 ext4_register_li_request(sb, first_not_zeroed);
4767 }
4768
6fd058f7 4769 ext4_setup_system_zone(sb);
d096ad0f 4770 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4771 ext4_commit_super(sb, 1);
0390131b 4772
ac27a0ec
DK		 4773#ifdef CONFIG_QUOTA
4774 /* Release old quota file names */
4775 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4776 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4777 if (enable_quota) {
4778 if (sb_any_quota_suspended(sb))
4779 dquot_resume(sb, -1);
4780 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4781 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4782 err = ext4_enable_quotas(sb);
07724f98 4783 if (err)
7c319d32 4784 goto restore_opts;
7c319d32
AK		 4785 }
4786 }
ac27a0ec 4787#endif
d4c402d9
CW		 4788
4789 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4790 kfree(orig_data);
ac27a0ec 4791 return 0;
0b8e58a1 4792
ac27a0ec
DK		 4793restore_opts:
4794 sb->s_flags = old_sb_flags;
4795 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4796 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4797 sbi->s_resuid = old_opts.s_resuid;
4798 sbi->s_resgid = old_opts.s_resgid;
4799 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4800 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4801 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4802#ifdef CONFIG_QUOTA
4803 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4804 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4805 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4806 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4807 }
4808#endif
d4c402d9 4809 kfree(orig_data);
ac27a0ec
DK		 4810 return err;
4811}
4812
2b2d6d01 4813static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4814{
4815 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4816 struct ext4_sb_info *sbi = EXT4_SB(sb);
4817 struct ext4_super_block *es = sbi->s_es;
27dd4385 4818 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4819 u64 fsid;
d02a9391 4820 s64 bfree;
27dd4385 4821 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4822
952fc18e
TT		 4823 if (!test_opt(sb, MINIX_DF))
4824 overhead = sbi->s_overhead;
ac27a0ec 4825
617ba13b 4826 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4827 buf->f_bsize = sb->s_blocksize;
b72f78cb 4828 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4829 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4830 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4831 /* prevent underflow in case that few free space is available */
57042651 4832 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4833 buf->f_bavail = buf->f_bfree -
4834 (ext4_r_blocks_count(es) + resv_blocks);
4835 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4836 buf->f_bavail = 0;
4837 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4838 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4839 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4840 fsid = le64_to_cpup((void *)es->s_uuid) ^
4841 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4842 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4843 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4844
ac27a0ec
DK		 4845 return 0;
4846}
4847
0b8e58a1
AD		 4848/* Helper function for writing quotas on sync - we need to start transaction
4849 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4850 * Process 1 Process 2
617ba13b 4851 * ext4_create() quota_sync()
a269eb18 4852 * jbd2_journal_start() write_dquot()
871a2931 4853 * dquot_initialize() down(dqio_mutex)
dab291af 4854 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4855 *
4856 */
4857
4858#ifdef CONFIG_QUOTA
4859
4860static inline struct inode *dquot_to_inode(struct dquot *dquot)
4861{
4c376dca 4862 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4863}
4864
617ba13b 4865static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4866{
4867 int ret, err;
4868 handle_t *handle;
4869 struct inode *inode;
4870
4871 inode = dquot_to_inode(dquot);
9924a92a 4872 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4873 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4874 if (IS_ERR(handle))
4875 return PTR_ERR(handle);
4876 ret = dquot_commit(dquot);
617ba13b 4877 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4878 if (!ret)
4879 ret = err;
4880 return ret;
4881}
4882
617ba13b 4883static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4884{
4885 int ret, err;
4886 handle_t *handle;
4887
9924a92a 4888 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4889 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4890 if (IS_ERR(handle))
4891 return PTR_ERR(handle);
4892 ret = dquot_acquire(dquot);
617ba13b 4893 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4894 if (!ret)
4895 ret = err;
4896 return ret;
4897}
4898
617ba13b 4899static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4900{
4901 int ret, err;
4902 handle_t *handle;
4903
9924a92a 4904 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4905 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4906 if (IS_ERR(handle)) {
4907 /* Release dquot anyway to avoid endless cycle in dqput() */
4908 dquot_release(dquot);
ac27a0ec 4909 return PTR_ERR(handle);
9c3013e9 4910 }
ac27a0ec 4911 ret = dquot_release(dquot);
617ba13b 4912 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4913 if (!ret)
4914 ret = err;
4915 return ret;
4916}
4917
617ba13b 4918static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4919{
262b4662
JK		 4920 struct super_block *sb = dquot->dq_sb;
4921 struct ext4_sb_info *sbi = EXT4_SB(sb);
4922
2c8be6b2 4923 /* Are we journaling quotas? */
262b4662
JK		 4924 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4925 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4926 dquot_mark_dquot_dirty(dquot);
617ba13b 4927 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4928 } else {
4929 return dquot_mark_dquot_dirty(dquot);
4930 }
4931}
4932
617ba13b 4933static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4934{
4935 int ret, err;
4936 handle_t *handle;
4937
4938 /* Data block + inode block */
9924a92a 4939 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 4940 if (IS_ERR(handle))
4941 return PTR_ERR(handle);
4942 ret = dquot_commit_info(sb, type);
617ba13b 4943 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4944 if (!ret)
4945 ret = err;
4946 return ret;
4947}
4948
4949/*
4950 * Turn on quotas during mount time - we need to find
4951 * the quota file and such...
4952 */
617ba13b 4953static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4954{
287a8095
CH		 4955 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4956 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4957}
4958
4959/*
4960 * Standard function to be called on quota_on
4961 */
617ba13b 4962static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4963 struct path *path)
ac27a0ec
DK		 4964{
4965 int err;
ac27a0ec
DK		 4966
4967 if (!test_opt(sb, QUOTA))
4968 return -EINVAL;
0623543b 4969
ac27a0ec 4970 /* Quotafile not on the same filesystem? */
d8c9584e 4971 if (path->dentry->d_sb != sb)
ac27a0ec 4972 return -EXDEV;
0623543b
JK		 4973 /* Journaling quota? */
4974 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4975 /* Quotafile not in fs root? */
f00c9e44 4976 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4977 ext4_msg(sb, KERN_WARNING,
4978 "Quota file not on filesystem root. "
4979 "Journaled quota will not work");
2b2d6d01 4980 }
0623543b
JK		 4981
4982 /*
4983 * When we journal data on quota file, we have to flush journal to see
4984 * all updates to the file when we bypass pagecache...
4985 */
0390131b 4986 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4987 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4988 /*
4989 * We don't need to lock updates but journal_flush() could
4990 * otherwise be livelocked...
4991 */
4992 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4993 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4994 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4995 if (err)
7ffe1ea8 4996 return err;
0623543b
JK		 4997 }
4998
f00c9e44 4999 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5000}
5001
7c319d32
AK		 5002static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5003 unsigned int flags)
5004{
5005 int err;
5006 struct inode *qf_inode;
5007 unsigned long qf_inums[MAXQUOTAS] = {
5008 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5009 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5010 };
5011
5012 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5013
5014 if (!qf_inums[type])
5015 return -EPERM;
5016
5017 qf_inode = ext4_iget(sb, qf_inums[type]);
5018 if (IS_ERR(qf_inode)) {
5019 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5020 return PTR_ERR(qf_inode);
5021 }
5022
bcb13850
JK		 5023 /* Don't account quota for quota files to avoid recursion */
5024 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5025 err = dquot_enable(qf_inode, type, format_id, flags);
5026 iput(qf_inode);
5027
5028 return err;
5029}
5030
5031/* Enable usage tracking for all quota types. */
5032static int ext4_enable_quotas(struct super_block *sb)
5033{
5034 int type, err = 0;
5035 unsigned long qf_inums[MAXQUOTAS] = {
5036 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5037 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5038 };
5039
5040 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5041 for (type = 0; type < MAXQUOTAS; type++) {
5042 if (qf_inums[type]) {
5043 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5044 DQUOT_USAGE_ENABLED);
5045 if (err) {
5046 ext4_warning(sb,
72ba7450
TT		 5047 "Failed to enable quota tracking "
5048 "(type=%d, err=%d). Please run "
5049 "e2fsck to fix.", type, err);
7c319d32
AK		 5050 return err;
5051 }
5052 }
5053 }
5054 return 0;
5055}
5056
5057/*
5058 * quota_on function that is used when QUOTA feature is set.
5059 */
5060static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5061 int format_id)
5062{
5063 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5064 return -EINVAL;
5065
5066 /*
5067 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5068 */
5069 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5070}
5071
ca0e05e4
DM		 5072static int ext4_quota_off(struct super_block *sb, int type)
5073{
21f97697
JK		 5074 struct inode *inode = sb_dqopt(sb)->files[type];
5075 handle_t *handle;
5076
87009d86
DM		 5077 /* Force all delayed allocation blocks to be allocated.
5078 * Caller already holds s_umount sem */
5079 if (test_opt(sb, DELALLOC))
ca0e05e4 5080 sync_filesystem(sb);
ca0e05e4 5081
0b268590
AG		 5082 if (!inode)
5083 goto out;
5084
21f97697
JK		 5085 /* Update modification times of quota files when userspace can
5086 * start looking at them */
9924a92a 5087 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5088 if (IS_ERR(handle))
5089 goto out;
5090 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5091 ext4_mark_inode_dirty(handle, inode);
5092 ext4_journal_stop(handle);
5093
5094out:
ca0e05e4
DM		 5095 return dquot_quota_off(sb, type);
5096}
5097
7c319d32
AK		 5098/*
5099 * quota_off function that is used when QUOTA feature is set.
5100 */
5101static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5102{
5103 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5104 return -EINVAL;
5105
5106 /* Disable only the limits. */
5107 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5108}
5109
ac27a0ec
DK		 5110/* Read data from quotafile - avoid pagecache and such because we cannot afford
5111 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5112 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5113 * we don't have to be afraid of races */
617ba13b 5114static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5115 size_t len, loff_t off)
5116{
5117 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5118 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5119 int err = 0;
5120 int offset = off & (sb->s_blocksize - 1);
5121 int tocopy;
5122 size_t toread;
5123 struct buffer_head *bh;
5124 loff_t i_size = i_size_read(inode);
5125
5126 if (off > i_size)
5127 return 0;
5128 if (off+len > i_size)
5129 len = i_size-off;
5130 toread = len;
5131 while (toread > 0) {
5132 tocopy = sb->s_blocksize - offset < toread ?
5133 sb->s_blocksize - offset : toread;
617ba13b 5134 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5135 if (err)
5136 return err;
5137 if (!bh) /* A hole? */
5138 memset(data, 0, tocopy);
5139 else
5140 memcpy(data, bh->b_data+offset, tocopy);
5141 brelse(bh);
5142 offset = 0;
5143 toread -= tocopy;
5144 data += tocopy;
5145 blk++;
5146 }
5147 return len;
5148}
5149
5150/* Write to quotafile (we know the transaction is already started and has
5151 * enough credits) */
617ba13b 5152static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5153 const char *data, size_t len, loff_t off)
5154{
5155 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5156 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5157 int err = 0;
5158 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5159 struct buffer_head *bh;
5160 handle_t *handle = journal_current_handle();
5161
0390131b 5162 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5163 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5164 " cancelled because transaction is not started",
9c3013e9
JK		 5165 (unsigned long long)off, (unsigned long long)len);
5166 return -EIO;
5167 }
67eeb568
DM		 5168 /*
5169 * Since we account only one data block in transaction credits,
5170 * then it is impossible to cross a block boundary.
5171 */
5172 if (sb->s_blocksize - offset < len) {
5173 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5174 " cancelled because not block aligned",
5175 (unsigned long long)off, (unsigned long long)len);
5176 return -EIO;
5177 }
5178
67eeb568
DM		 5179 bh = ext4_bread(handle, inode, blk, 1, &err);
5180 if (!bh)
5181 goto out;
62d2b5f2
JK		 5182 err = ext4_journal_get_write_access(handle, bh);
5183 if (err) {
5184 brelse(bh);
5185 goto out;
ac27a0ec 5186 }
67eeb568
DM		 5187 lock_buffer(bh);
5188 memcpy(bh->b_data+offset, data, len);
5189 flush_dcache_page(bh->b_page);
5190 unlock_buffer(bh);
62d2b5f2 5191 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5192 brelse(bh);
ac27a0ec 5193out:
0b7f7cef 5194 if (err)
ac27a0ec 5195 return err;
67eeb568
DM		 5196 if (inode->i_size < off + len) {
5197 i_size_write(inode, off + len);
617ba13b 5198 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5199 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5200 }
67eeb568 5201 return len;
ac27a0ec
DK		 5202}
5203
5204#endif
5205
152a0836
AV		 5206static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5207 const char *dev_name, void *data)
ac27a0ec 5208{
152a0836 5209 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5210}
5211
37f328eb 5212#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5213static inline void register_as_ext2(void)
5214{
5215 int err = register_filesystem(&ext2_fs_type);
5216 if (err)
5217 printk(KERN_WARNING
5218 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5219}
5220
5221static inline void unregister_as_ext2(void)
5222{
5223 unregister_filesystem(&ext2_fs_type);
5224}
2035e776
TT		 5225
5226static inline int ext2_feature_set_ok(struct super_block *sb)
5227{
5228 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5229 return 0;
5230 if (sb->s_flags & MS_RDONLY)
5231 return 1;
5232 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5233 return 0;
5234 return 1;
5235}
24b58424
TT		 5236#else
5237static inline void register_as_ext2(void) { }
5238static inline void unregister_as_ext2(void) { }
2035e776 5239static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5240#endif
5241
37f328eb 5242#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5243static inline void register_as_ext3(void)
5244{
5245 int err = register_filesystem(&ext3_fs_type);
5246 if (err)
5247 printk(KERN_WARNING
5248 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5249}
5250
5251static inline void unregister_as_ext3(void)
5252{
5253 unregister_filesystem(&ext3_fs_type);
5254}
2035e776
TT		 5255
5256static inline int ext3_feature_set_ok(struct super_block *sb)
5257{
5258 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5259 return 0;
5260 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5261 return 0;
5262 if (sb->s_flags & MS_RDONLY)
5263 return 1;
5264 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5265 return 0;
5266 return 1;
5267}
24b58424
TT		 5268#else
5269static inline void register_as_ext3(void) { }
5270static inline void unregister_as_ext3(void) { }
2035e776 5271static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5272#endif
5273
03010a33
TT		 5274static struct file_system_type ext4_fs_type = {
5275 .owner = THIS_MODULE,
5276 .name = "ext4",
152a0836 5277 .mount = ext4_mount,
03010a33
TT		 5278 .kill_sb = kill_block_super,
5279 .fs_flags = FS_REQUIRES_DEV,
5280};
7f78e035 5281MODULE_ALIAS_FS("ext4");
03010a33 5282
8f021222 5283static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5284{
5285 struct ext4_features *ef;
5286 int ret = -ENOMEM;
5287
5288 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5289 if (!ef)
5290 goto out;
5291
5292 ef->f_kobj.kset = ext4_kset;
5293 init_completion(&ef->f_kobj_unregister);
5294 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5295 "features");
5296 if (ret) {
5297 kfree(ef);
5298 goto out;
5299 }
5300
5301 ext4_feat = ef;
5302 ret = 0;
5303out:
5304 return ret;
5305}
5306
8f021222
LC		 5307static void ext4_exit_feat_adverts(void)
5308{
5309 kobject_put(&ext4_feat->f_kobj);
5310 wait_for_completion(&ext4_feat->f_kobj_unregister);
5311 kfree(ext4_feat);
5312}
5313
e9e3bcec
ES		 5314/* Shared across all ext4 file systems */
5315wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5316struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5317
5dabfc78 5318static int __init ext4_init_fs(void)
ac27a0ec 5319{
e9e3bcec 5320 int i, err;
c9de560d 5321
07c0c5d8
AV		 5322 ext4_li_info = NULL;
5323 mutex_init(&ext4_li_mtx);
5324
9a4c8019 5325 /* Build-time check for flags consistency */
12e9b892 5326 ext4_check_flag_values();
e9e3bcec
ES		 5327
5328 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5329 mutex_init(&ext4__aio_mutex[i]);
5330 init_waitqueue_head(&ext4__ioend_wq[i]);
5331 }
5332
51865fda 5333 err = ext4_init_es();
6fd058f7
TT		 5334 if (err)
5335 return err;
51865fda
ZL		 5336
5337 err = ext4_init_pageio();
5338 if (err)
5339 goto out7;
5340
5dabfc78 5341 err = ext4_init_system_zone();
bd2d0210 5342 if (err)
d44651d0 5343 goto out6;
3197ebdb 5344 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5345 if (!ext4_kset) {
5346 err = -ENOMEM;
dd68314c 5347 goto out5;
0e376b1e 5348 }
d44651d0 5349 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5350
5351 err = ext4_init_feat_adverts();
dd68314c
TT		 5352 if (err)
5353 goto out4;
857ac889 5354
5dabfc78 5355 err = ext4_init_mballoc();
ac27a0ec 5356 if (err)
6fd058f7 5357 goto out3;
c9de560d 5358
5dabfc78 5359 err = ext4_init_xattr();
c9de560d
AT		 5360 if (err)
5361 goto out2;
ac27a0ec
DK		 5362 err = init_inodecache();
5363 if (err)
5364 goto out1;
24b58424 5365 register_as_ext3();
2035e776 5366 register_as_ext2();
03010a33 5367 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5368 if (err)
5369 goto out;
bfff6873 5370
ac27a0ec
DK		 5371 return 0;
5372out:
24b58424
TT		 5373 unregister_as_ext2();
5374 unregister_as_ext3();
ac27a0ec
DK		 5375 destroy_inodecache();
5376out1:
5dabfc78 5377 ext4_exit_xattr();
c9de560d 5378out2:
5dabfc78 5379 ext4_exit_mballoc();
6fd058f7 5380out3:
8f021222 5381 ext4_exit_feat_adverts();
dd68314c 5382out4:
d44651d0
FJ		 5383 if (ext4_proc_root)
5384 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5385 kset_unregister(ext4_kset);
d44651d0 5386out5:
5dabfc78 5387 ext4_exit_system_zone();
d44651d0 5388out6:
5dabfc78 5389 ext4_exit_pageio();
51865fda
ZL		 5390out7:
5391 ext4_exit_es();
5392
ac27a0ec
DK		 5393 return err;
5394}
5395
5dabfc78 5396static void __exit ext4_exit_fs(void)
ac27a0ec 5397{
bfff6873 5398 ext4_destroy_lazyinit_thread();
24b58424
TT		 5399 unregister_as_ext2();
5400 unregister_as_ext3();
03010a33 5401 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5402 destroy_inodecache();
5dabfc78
TT		 5403 ext4_exit_xattr();
5404 ext4_exit_mballoc();
8f021222 5405 ext4_exit_feat_adverts();
9f6200bb 5406 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5407 kset_unregister(ext4_kset);
5dabfc78
TT		 5408 ext4_exit_system_zone();
5409 ext4_exit_pageio();
ac27a0ec
DK		 5410}
5411
5412MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5413MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5414MODULE_LICENSE("GPL");
5dabfc78
TT		 5415module_init(ext4_init_fs)
5416module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository